Multiple View Discriminative Appearance Modeling with IMCMC for Distributed Tracking

Size: px
Start display at page:

Download "Multiple View Discriminative Appearance Modeling with IMCMC for Distributed Tracking"

Transcription

1 Muliple View Discriminaive ing wih IMCMC for Disribued Tracking Sanhoshkumar Sunderrajan, B.S. Manjunah Deparmen of Elecrical and Compuer Engineering Universiy of California, Sana Barbara Absrac This paper proposes a disribued muli-camera racking algorihm wih ineracing paricle filers. A robus muli-view appearance model is obained by sharing raining samples beween views. Moivaed by incremenal learning and [1], we creae an inermediae daa represenaion beween wo camera views wih generaive subspaces as poins on a Grassmann manifold, and sample along he geodesic beween raining daa from wo views o uncover he meaningful descripion due o viewpoin changes. Finally, a Boosed appearance model is rained using he projeced raining samples on o hese generaive subspaces. For each objec, a se of wo paricle filers i.e., local and global is used. The local paricle filer models he objec moion in he image plane. The global paricle filer models he objec moion in he ground plane. These paricle filers are inegraed ino a unified Ineracing Markov Chain Mone Carlo (IMCMC) framework. We show he manner in which we induce priors on scene specific informaion ino he global paricle filer o improve racking accuracy. The proposed algorihm is validaed wih exensive experimenaion in challenging camera nework daa, and compares favorably wih sae of he ar objec rackers. Keywords Disribued Camera Nework, Paricle Filers, Objec Tracking. Predic and Updae Local Paricle Filer wih local informaion. Predic and Updae Local Paricle Filer wih local ineracions. Predic and Updae Ineracion Global Paricle Filer wih Muli-camera informaion and Scene Priors. Updae Local and Global Appearance Camera 1 Camera N Predic and Updae Global Paricle Filer wih Muli-camera informaion and Scene Priors. Fig. 1: Proposed Muli-camera based racking algorihm: Boh local and global paricle filers operae in parallel. Local paricle filer ineracs wih he global paricle filer using IMCMC framework [6]. Local paricle filer akes only local informaion available wihin he camera view ino accoun. Global paricle akes muli-camera informaion (performs fusion) and scene priors ino accoun. Finally, a robus global appearance model is learn by sharing raining samples wih he neighboring camera views. Conribuions of his paper are highlighed in red. I. INTRODUCTION Proliferaion of cheap and nework enabled smar cameras has provided an enormous opporuniy for large scale deploymen of camera neworks in real-world applicaions. This paper proposes a novel racking algorihm o rack pedesrians using a se of cameras wih overlapping views by inegraing informaion from differen camera views. The ground plane homography is used o relae informaion obained from muliple cameras o esimae an objec s posiion in he ground plane. Furher, o enhance racking qualiy, scene priors such as crowd flow informaion from he pas hisorical daa is used. More imporanly, we propose a sraegy o learn a robus discriminaive appearance model by sampling raining examples from generaive subspaces beween wo views and his leads o significan improvemen in racking accuracy. Exising muli-camera racking algorihms assume ha he racking problem in individual cameras is solved, and focus on higher level asks such as aciviy recogniion, even analysis and camera-hand-off. Due o background disorions, complex illuminaion changes, varying objec shapes and packe losses in wireless communicaion, he performance of single camera racking is far from he ideal. Also, he racking algorihm is mos likely o fail due o improper appearance modeling. Muliple viewpoin changes are no explicily modeled in exising racking algorihms. More imporanly, informaion fusion algorihms do no ake scene priors ino accoun. There are a number of muli-camera racking algorihms [2], ye very lile aenion has been paid o disribued racking algorihms. Exising disribued racking algorihms [3], [4], [5] manifes as soluions o an informaion fusion problem and do no ake prior informaion abou he nework ino consideraion during he fusion. Also, no much aenion has been paid o robus muli-camera appearance modeling. We propose a novel sraegy o rain a discriminaive Boosed appearance model by sharing raining samples wih neighboring views by aking view-shif ino accoun. More imporanly, we induce he priors on scene informaion such as crowd flow ino a paricle filer framework and demonsrae ha his significanly improves he robusness of objec racking. For a synchronized nework wih M cameras, le {I c 1:} be a se of video frames from differen camera views where c {1... M}. A ime insance, he sae of he i h objec on he image plane of he c h camera is denoed by X c,i = [posiionx, posiiony, sizex, sizey ] where i {1,..., O} is he global objec index and O is he number of objecs iniialized by he objec deecor a he firs frame. [posiionx, posiiony ] is he cener of he objec s bounding

2 box on he image plane and [sizex, sizey ] is he size of he objec along he x and y direcions. Given he se of video frames, i.e. {I c 1:} from differen cameras, we infer he objec sae for he i h objec, {X c,i } = [X 1,i... X M,i ] by Maximum a Poseriori (MAP) formulaion: arg max p({x c,i {X c,i } {I c 1:}) (1) } where i is he objec index and c is he camera index. We propose a novel sraegy o rain a robus discriminaive muli-view appearance model p(y c,i X c,i ), by aking viewpoin changes ino accoun. Addiionally, we propose a unified Markov Chain Mone Carlo (MCMC) framework o combine local and global informaion ino he racker. Mos imporanly, we show he manner in which we induce he scene specific priors and mulicamera ineracion ino he global paricle filer o improve he racking accuracy. Figure 1 highlighs he conribuions of he proposed muli-camera racking algorihm. Following are he main conribuions of his paper: 1. A robus globally discriminaive appearance modeling by aking viewpoin shif ino accoun. 2. A unified MCMC approach o combine local and global models. Global paricle filer models he muli-camera informaion fusion and akes scene priors ino accoun. II. RELATED WORKS In [7], a disribued Kalman filering framework is used o rack he objecs on he global ground plane. There have been recen effors o perform acive collaboraion beween cameras. [8] proposed a Bayesian algorihm for disribued muli-arge racking using muliple collaboraive cameras and hey used 2D locaion insead of 3D. They do no ake higher level scene informaion ino accoun while performing he fusion. Especially in [9], a disribued fusion mechanism ha clusers paricles obained from muliple camera views is proposed. This mehod is highly vulnerable o ouliers in shared paricles. In conras, our disribued racking algorihm combines local and global moion models in a unified probabilisic framework. Also, scene priors are aken ino accoun while updaing he global moion model. There are a few single camera racking algorihms ha ake scene priors ino accoun o improve he racking accuracy, however, hese mehods are no sraighforward o exend o muli-camera disribued racking scenarios [10], [11]. Recenly, racking by deecion algorihms have been gaining populariy [12], [13], [14]. Exising muliple camera racking algorihms do no discriminaively model he muli-view appearance in an online manner. Roh e al. [15] proposed a muliple insance learning based co-raining sraegy for muliview appearance racking. In comparison, we propose a novel disribued sraegy o rain a discriminaive appearance model by aking viewpoin shif ino accoun. The res of he paper is organized as follows. Secion III inroduces Bayesian Tracking formulaion wih Markov Chain and Mone Carlo. Secion IV formulaes he mulicamera racking algorihm. Secion IV-A proposes a novel mehod for muli-view appearance modeling and secion IV-B discusses muli-camera informaion fusion by inducing scene priors. Secion V demonsraes he efficacy of he proposed mehodology on some challenging muli-camera daases and finally we conclude he paper in secion VI. III. BAYESIAN TRACKING FORMULATION We explain he basics of paricle filer based objec racking [11]. The goal of objec racking is o find he bes objec configuraion X given he observaions Y 1: up o ime using he following Bayesian formulaion: p(x Y 1: ) p(y X ) p(x X 1 )p(x 1 Y 1: 1 ) dx 1, The opimal objec configuraion ˆX is obained by Maximum a Poseriori (MAP) esimaion: (2) ˆX = arg max X p(x Y 1: ) (3) The poserior a ime 1, is approximaed by a se of weighed paricles in Sequenial Imporance Resampling (SIR) paricle filers: p(x 1 Y 1 ) {X (p) 1, π(p) 1 }P p=1 (4) where p is he paricle index and P is he number of paricles. The weigh of he p h paricle is given by, π (p) = p(y X (p) ). The inegral in he equaion 2 can be approximaed by: p(x Y ) p(y X ) P p=1 π (p) 1 p(x X (p) 1 ) (5) We use MCMC sampling insead of he sandard imporance resampling. Compared o imporance resampling based paricle filer, he MCMC based paricle filer is very efficien [16]. The complexiy varies linearly wih respec o number of objecs compared o exponenially varying complexiy in SIR paricle filers. More imporanly, imporance resampling mehods suffer from paricle impoverishmen and degeneracy. In MCMC mehods, a Markov chain is defined over he space of configuraions X and he saionary disribuion of he chain is equal o he poserior disribuion, p(x Y). A se of un-weighed samples i.e., p(x Y ) {X (p) } P p=1 is used o represen he poserior in MCMC based paricle filer. IV. DISTRIBUTED MULTI-OBJECT TRACKING IN A CAMERA NETWORK For a camera nework wih M synchronized cameras, he primary ask is o joinly rack objecs wih acive collaboraion beween he views. The goal is o esimae he sae of objecs, i.e, {X c,i } given he se of image observaions {Y c,i 1: } on he se of video frames {I c 1:} from differen cameras and ground plane homography. We assume ha no raw image daa is ransferred beween nodes due o nework bandwidh consrains and esimae he following by approximaing equaion 1 on camera c = C for he i h objec as:

3 Updae Local Updae Global Share Training Samples Updae Global Updae Local Paricle Sae Predicion for local Filer from ime -1 Paricle Sae Predicion for global Filer from ime -1 nework channel Paricle Sae Predicion for global Filer from ime -1 Paricle Sae Predicion for local Filer from ime -1 Updae global paricle filer sae a ime Updae global paricle filer sae a ime Updae local paricle filer sae a ime Share Objec Measuremen INTERACTION Camera 1 Camera N INTERACTION Updae local paricle filer sae a ime Trajecory Daabase Trajecory Daabase Fig. 2: Ineracive MCMC based racking: A ime 1, he global appearance model is rained by sharing raining examples beween differen views. The local appearance model is rained using examples wihin he image view. Boh local and global paricle filers predic he sae from ime 1. The global paricle updaes is sae using he objec measuremen shared beween differen views. The local paricle filer ineracs wih he global paricle filer abou he bes configuraion of he objec. arg max p({x c,i {X c,i } I c=c 1: ), c = 1... M (6) } By assuming I c=c is condiionally independen of I c=c 1: 1 given he esimaes {X c,i } and expanding equaion 6 by Bayes rule, we have he following: p({x c,i } I c=c 1: ) p(i c=c {X c,i })p({x c,i } I1: 1) c=c (7) On furher facorizing he image likelihood erm }), we ge: p(i c=c {X c,i p({x c,i } I c=c 1: ) Appearance Likelihood p(y c=c,i X c=c,i ) P rior p({x c,i } I c=c 1: 1) Spaial Likelihood p({x c C,i } X c=c,i ) We propose a novel sraegy o model he appearance likelihood in a disribued manner by aking viewpoin shif ino accoun (explained in secion IV-A). Also, in secion IV-B we show he manner in which we combine local and global moion models using he ineracing Markov Chain Mone Carlo. Mos imporanly, we induce scene specific priors ino he global paricle filer. Figure 2 shows he proposed disribued racking mehodology. A. ing We use a discriminaive classifier o model he appearance of he objec. The likelihood of a pixel x belonging o he foreground label y = 1 is given by: 1 p(y = 1 x) = (9) 1 + e H(F(x)) where H is a discriminaive classifier learn wih online Boosing and F is he image feaure compued a pixel locaion x. Given a confidence map of he objec of ineres by esing appearance classifiers a ime, a mean-shif based [12] objec (8) sae esimae ν is obained. The appearance likelihood is given as follows: p(y X ) = p pos (ν X ) (10) where p pos (ν X ) is drawn from he Normal disribuion such ha p pos (ν X ) N (ν ; X, Σ pos ) and Σ pos is he covariance marix. In he res of his secion, we discuss in deail abou he manner in which we rain he local and global appearance classifiers, H (l) and H (g) respecively. 1) Local ing: For learning he local appearance classifier H (l), we use a combinaion of hisogram of oriened gradiens feaures (HOG) [17] and normalized color feaures. We use a 12 dimensional feaure vecor wih 9 bins for HOG and 3 for normalized pixel RGB color values. Similar o Ensemble Tracker [12], we rain a discriminaive online classifier a each ime insance by using samples wihin he objec bounding box as posiive examples and samples ouside he objec bounding box as negaive examples. 2) Global ing: Similar o local appearance modeling, we exrac an n-dimensional feaure vecor wih HOG and normalized pixel RGB color values (we used 36 bins for HOG and 3 for normalized pixel RGB color values). We propose a novel sraegy o exrac raining samples for learning he global discriminaive appearance model in a disribued manner. A each ime insance, each racker in every view (one racker per objec) shares heir raining examples o heir neighboring camera views. We firs formulae he problem for wo views and subsequenly exend o muliple views. Le X 1 R N1 n and X 2 R N2 n be he se of raining samples obained from wo views where N 1 and N 2 are he number of raining samples in each of he views. S 1 R N1 d and S 2 R N2 d represen generaive subspaces obained by performing Principal Componen Analysis (PCA) on X 1 and X 2 where d << n (in our experimens, we se d = 17). We now discuss he compuaion of inermediae subspaces S m, m R, 1 < m < 2 in order o model he appearance changes due o viewpoin variaions. The space of d-dimensional subspaces conaining he origin in R n can be idenified wih he Grassmann manifold G n,d where S 1 and S 2 are poins on G n,d. The Grassmann manifold is he space of d dimensional subspace in R n and a poin on Grassmann manifold represens a subspace. We use geodesic pahs ha are consan velociy curves on a manifold o obain inermediae subspaces. Viewing G n,d as quoien space of SO(n), he geodesic pah saring from S 1 is given by oneparameer exponenial flow ψ(m ) = Qexp(m B) [1], [18] where 0 < m < 1 and B is resriced o a skew-symmeric, block diagonal marix of he form: [ ] 0 A T B =, A R A 0 (n d) d. (11) where A specifies he direcion and he speed of [ geodesic ] flow. Q SO(n) such ha Q T Id S 1 = J and J =. I 0 d is n d,d a d d ideniy marix. By varying m beween 0 and 1, we ge a se of inermediae subspace S ha includes S 1 and S 2 (ineresed readers refer o [1] for more deails on calculaing A).

4 . HOG HOG View 1 View 2 G n,d ihog ihog p l (Y c=c,i X c=c,i ) = ψ(x c=c,i Appearance wih H (l) p(y c=c,i X c=c,i ) Ineracion { }}{ ψ(x c=c,i, X c=c,j ) j i (12), X c=c,j ) encodes he ineracion likelihood based on he sae of he oher objecs in he scene and i is based on he Magneic repulsion poenial: m' = 0.2 m' = 0.45 m' = 0.55 m' = 0.75 Fig. 3: Training Sample Generaion: Training samples for global appearance learning is obained by projecing samples from view 1 on o differen generaive subspaces obained by varying m. As highlighed in inverse HOG represenaion (ihog) obained using [19], body of he person is clearly discriminaive wih respec o viewpoin changes. Given he inermediae subspaces S, we propose a novel mehodology o exrac raining samples o learn he muliview discriminaive appearance model. Le N be he number of inermediae subspaces. We exrac raining samples by projecing X 1 on o hese inermediae subspaces. We rain our global discriminaive appearance model by gahering each of he raining samples projeced on o hese inermediae subspaces i.e., X 1 R N N1 d. For muliple views, we adop similar sraegy o exrac raining samples beween he curren view and every oher view using inermediae subspaces. Finally, we random sample raining examples o learn he global appearance classifier. B. Ineracing Markov Chain Mone Carlo Based Tracking We now presen a camera racking-by-deecion algorihm wih an ineracing MCMC framework. For every objec, we use wo kinds of paricle filers, local and global, wih MCMC sampling. For local paricle filers, he observaion likelihood is compued using he local appearance model and objec ineracion ha are local o he camera. For global paricle filers, he observaion likelihood is compued based on global appearance model, muli-camera informaion and scene priors. Similar o [20], each filer operaes eiher in parallel or ineracive mode. In he parallel mode, objec sae is compleely deermined by he local paricle filer. In he ineracing mode, he local paricle filer ineracs wih he global filer and deermines he objec sae a ha ime insance. By separaing he local and global models, we can easily accoun for higher level consrains ino he global model and reduce he complexiy of he local paricle filers significanly. C. MCMC based Local Paricle Filer The local paricle filer capures he local informaion wihin he image plane and i does no ake scene priors ino accoun for modeling he objec moion. The observaion model for he local paricle filer is given by: ψ(x c=c,i, X c=c,j ) = 1 1 exp( dis(xc=c,i β σ 2 i, X c=c,j ) ) (13) where β is he normalizaion consan and σi 2 characerizes he allowable maximum ineracion disance (in our experimens, we se β = 1 and σi 2 = 100). We approximae he local moion model wih he Normal disribuion such ha p l (X X 1 ) N (X ; X 1, Σ l ). The local paricle filer finds he MAP esimae defined in equaion 3 by sampling via Meropolis Hasings algorihm. The algorihm consiss of wo seps, a proposal sep and an accepance sep. In he proposal sep, he new sae is proposed wih he following proposal densiy: R l (X ; X ) = p l (X X ) (14) where R l denoes he proposal densiy funcion based on he local paricle filer s moion model and X represens he new sae proposed by R l a ime. Given he proposed sae, he local filer acceps he new sae wih he accepance raio given by: [ α parallel = min 1, p l(y X )R l (X ; X ) ] p l (Y X )R l (X ; X ) D. MCMC based Global Paricle Filer (15) The global paricle filer operaes wih he global appearance model and i akes scene priors and muli-camera informaion ino accoun. The observaion model, p g (Y c=c,i X c=c,i ), for he global paricle filer is given by: p g (Y c=c,i X c=c,i ) = Appearance wih H (g) p(y c=c,i X c=c,i ) Muli camera likelihood { }}{ Γ(X c=c,i, X c=k,i ) k C SceneP riors φ(x c=c,i ) (16) Γ(X c=c,i, X c=k,j ) encodes he muli-camera spaial likelihood and i is given by: Γ(X c=c,i, X c=k,i ) = 1 exp( dis(xc=c,i ρ σ 2 c, X c=k,i ) ) (17)

5 TABLE I: Mean Roo Mean Square Pixel Errors on differen daases. Daases OAB OAB-PF MS MIL MIL-PF Sruck Proposed Oudoor PETS Indoor TABLE II: Mean Pascal VOC deecion scores on differen daases. Daases OAB OAB-PF MS MIL MIL-PF Sruck Proposed Oudoor PETS Indoor where ρ is he normalizaion consan and σc 2 characerizes he allowable disance for muli-camera ineracion (in our experimens, we se ρ = 1 and σc 2 = 1000). We share objec s measuremen (mean-shif esimae obained by global appearance classifier) wih he neighboring camera views for mulicamera ineracion. φ(x c=c,i ) is based on he scene, i encodes he scene knowledge ino he observaion model. For modeling he scene specific priors, we used Kernel densiy esimaor o model he probabiliy densiy funcion on rajecories of he objecs ha moved over he scene for a period of ime [10]. We approximae he global moion model wih he Normal disribuion such ha p g (X X 1 ) N (X ; X 1, Σ g ). Similar o he local paricle filer, he global paricle filer uses he Meropolis Hasings algorihm based MCMC sampling. Algorihm 1: Muli-Objec Tracking wih IMCMC Inpu: X, γ Oupu: ˆX 1: rand() reurns a random number beween 0 and 1. 2: if rand() < γ hen Accep he new sae wih he probabiliy (18) else Propose he new sae using (14) Accep he new sae wih he probabiliy (15) end 3: Esimae he MAP sae ˆX using (3) E. IMCMC based Tracking Algorihm During he sampling process, a each ime sep, he local paricle filer ineracs wih he global paricle filer (illusraed in Fig.2). We make use of he IMCMC framework for he local paricle filer o communicae wih he global paricle filer [6], [11]. However, he global paricle filer operaes enirely in parallel mode. The proposed racking algorihm operaes in eiher parallel or ineracive mode [20]. The local and global paricle filers ac as he parallel Meropolis Hasings in he parallel mode. Whereas in he ineracion mode, he local paricle filer communicaes wih he global paricle filer and seeks a beer sae for he objec configuraion. The local paricle filer hen acceps he sae of he global paricle filer as is own sae wih he probabiliy as given by: α ineracing = p l (Y c=c,i p g (Y c=c,i X c=c,i X c=c,i ) ) + p g (Y c=c,i X c=c,i ) (18) F. Occlusion Handling In some scenarios, objecs migh no be deeced due o various reasons such as lighing changes, illuminaion effecs, and missing feaures. We assume ha an objec is occluded when ζ falls below a cerain hreshold (0.1 in our experimens), where ζ of he image pach posiioned a X is given by: ζ = sum of pixel likelihoods area of he pach posiioned a X For a given racker, if he objec is occluded, he global paricle filer proceeds wih he predicion sep and performs an updae wih muli-camera informaion and scene priors. On he oher hand, he local paricle filer operaes compleely in ineracive mode. The racker pauses is operaion afer a specified number of coninuously missed deecions beween frames (in our experimens, we se he hreshold o 20 frames). V. EXPERIMENTS We evaluaed he proposed mehod wih a wide area camera nework consising of six cameras on an oudoor environmen. Videos (640x480) are capured for several hours in an unconrolled environmen wih complex shape and appearance changes in human body, wireless packe losses and irregular illuminaion variaions (for example shadows, lighing changes due o sun rays and ohers). Also, we evaluaed he proposed algorihm in some of he publicly available muli-camera pedesrian daases [21] and [9]. Tables I and II show average VOC deecion scores and roo mean square pixel errors for various algorihms on hree differen daases (Oudoor, PETS2009, and Indoor). In all he experimens, we se he number of weak classifiers for he Ensemble raining o 12 and updaed 2 weak classifiers a every frame. We se he number of paricles P = 500. For boh local and global paricle filers, we used he Brownian moion model wih sandard deviaion σ x = 21, σ y = 3 and σ s = We se he IMCMC ineracion hreshold γ = 0.1. We assumed Σ pos o be a diagonal marix wih leading diagonals equal o [2, 2]. Finally, for generaing inermediae subspaces we se m = [0.25, 0.5, 0.75]. For comparison merics, we used Roo Mean Square Pixel error and pascal VOC deecion score (primarily used for comparing single camera appearance based rackers). VOC deecion scores capure compacness of he prediced bounding box is wih respec o he ground plane bounding box. Wih he prediced bounding box B p and he ground ruh bounding box B g, he pascal VOC deecion score is given by: Algorihm 1 explains he proposed muli-camera racking mehodology wih IMCMC framework. The MAP esimae is obained using he local paricle filer by communicaing wih he global paricle filer: V OC Deecion Score = area(b p B g ) area(b p B g ) (19)

6 C2 C2 C2 C2 C1 C3 C6 O2 Proposed(Frame-45) OAB(Frame-45) MIL(Frame-45) Sruck(Frame-45) C4 C4 C4 C4 O2 Proposed(Frame-45) OAB(Frame-45) MIL(Frame-45) Sruck(Frame-45) Fig. 4: Experimen 1: On camera C2, objec is indisinguishable from he background and hence some of he discriminaive appearance model based rackers fail. Whereas in he proposed mehodology, raining samples are exraced from inermediae subspaces beween muliple views o learn he global appearance model. Addiionally, muli-camera informaion and scene priors are helpful in maneuvering he objec when appearance cues are misleading. Bes viewed in color. A. Global ing wih Global Filering In our firs se of experimens wih our own daase, we evaluaed he proposed mehodology wih six cameras in he nework, of which only wo of he cameras (C2 and C4) direcly sensed he objecs of ineres ( and O2). The objecs of ineres are iniialized using a background subracion based blob deecor. We compared he proposed mehodology wih some of he sae of he ar racking algorihms: on-line Adaboos (OAB) [22], on-line Adaboos wih paricle filer (OAB-PF), Muliple insance learning(mil) [13], Muliple insance learning wih paricle filer (MIL-PF), Sruck [23], and Mean-shif (MS) [24] rackers. From he oudoor scenario resuls in Tables I and II, we show ha he proposed mehodology ouperforms all he oher racking mehodologies in nework wide scenarios. Poor performance of some of he appearance based rackers is due o misaken ideniies and inadequae appearance modeling. In he case of he proposed algorihm, a robus global appearance classifier is learn by aking viewpoin shif ino accoun. Addiionally, he mulicamera likelihood and scene priors in global paricle filers are helpful in correcing he local paricle filers whenever he local paricle filer suffers from insufficien moion and appearance modeling. When he objec merges wih background (shown in Figure 4), scene priors in global paricle filers are helpful in maneuvering he objec using he prior informaion from previous objecs ha moved on he same locaion over he ime and also he muli-camera likelihood is helpful in correcing he objec sae esimae by exploiing muli-view redundancy. Mos imporanly, he global appearance model capures he missing feaures using he raining samples obained from neighboring camera views and helps he racker o recover iself in he subsequen frames by building a robus appearance model. B. Global ing wih Scene Priors For he second se of experimens wih PETS-2009 sequences, he objec of ineres () was iniialized in he firs (C1), hird (C3) and sixh (C6) views respecively. For his experimen, o sudy he efficacy of he global appearance modeling wih scene priors, he ineracive likelihood in local paricle filer was urned off. As seen in Tables I and II, he proposed algorihm ouperforms sae of he ar algorihms on VOC deecion scores and roo mean square pixel errors. The success of he proposed algorihm (as illusraed in Figure Fig. 5: Experimen 2: Tracking resuls from individual camera views are shown. Clearly, muli-camera informaion and scene priors help in avoiding failures due o parial occlusions. In his experimen, mos of he oher racking mehods failed due o he parial occlusion caused by he lamp pole. Objec of ineres is marked by he label. Bes viewed in color. 5) on his scenario could be aribued o he following: The proposed algorihm is robus o inermien racker failures due o he parial occlusions by learning a robus global appearance model. In his scenario, he lamp pole caused a parial occlusion in view C1 and mos of he racking algorihms failed compleely due o improper appearance modeling. Similar o he firs se of experimens, scene priors and muli-camera informaion is helpful in correcing he local paricle filers during insufficien moion modeling. C. Global ing wihou Scene Priors For he hird se of experimens, we used indoor video sequences from [9]. The camera nework consiss of five cameras along a long corridor. In order o increase he difficuly of he experimen, we used a par of he sequence where he corridor lighs were swiched off. Also, scene priors were no available for his nework. As seen in Tables I and II, he proposed algorihm ouperforms he res due o he following reasons: a) Objecs appeared similar due o he lighing condiions, he global appearance model is helpful in capuring discriminaive feaures using he raining samples from neighboring views. Oher mehods fail o capure hese variaions and hence lose rack of he objecs due o improper appearance based associaion, b) Also, he ineracive likelihood in he local paricle filer is helpful in mainaining spaial relaionship beween he objecs beween he frames. This helps in proper associaion in he presence of ouliers due o missing feaures beween he frames. Missing feaures are more likely o happen due o irregular lighing condiions. D. Comparison Wih Muliple Camera Trackers We compared he proposed algorihm wih sae of he ar muliple camera disribued filering based racking algorihms: Join Probabilisic Daa associaion wih Kalman consensus filer (JPDA-KCF) [25], [26], Informaion consensus filering wih neares-neighbor daa associaion (ICF-NN) [5], Informaion consensus filering wih ground ruh daa associaion (ICF-GT) and Muli-camera informaion consensus (MTIC) [3]. In ICF-NN, he neares observaion is associaed wih he exising rackle based on Hungarian algorihm. We used Sruck [23] (an appearance based racker) o generae image based observaions. Ground plane esimaes are obained using homographic ransformaion and hey serve as measuremens for ground disribued fusion algorihms. Tables III show mean error (in meers) and error sandard deviaion for differen algorihms in oudoor/indoor sequences respecively. For he indoor sequences, scene priors were no available. The proposed algorihm clearly ouperforms oher muli-camera racking algorihms due o he following: a)

7 TABLE III: Muliple Camera Tracking Comparison in Oudoor/Indoor Sequences (Muliple Objecs) Algorihm Mean Error (m) Error Sandard Deviaion (m) Proposed 0.27 / / 0.39 MTIC 6.40 / / 0.44 ICF-NN 25.7 / / 0.41 JPDA-KCF 52.5 / / 0.46 ICF-GT 12.8 / / 0.40 Muli-view appearance model effecively capures viewpoin variaions. b) Ground plane fusion in global paricle filers is efficienly fed back o he local paricle filers hrough an IMCMC approach. VI. CONCLUSION This paper presened a robus muli-camera racking algorihm using ineracing Markov Chain Mone Carlo. The racking algorihm is formulaed as a global Bayesian esimaion problem and solved in a disribued manner. We proposed a novel algorihm o learn a discriminaive muli-view appearance model by sharing samples across he views. We provided an efficien algorihm o combine local and global models ino one using a unified probabilisic framework. The disribued muli-camera racker akes approximaely one second perframe wih he Malab implemenaion on a machine wih 8 GB RAM and 2.67 GHZ processor. The proposed algorihm is esed on some challenging daases and validaed wih objecive resuls. As a fuure work, we plan o add complex crowd behavioral model ino he local objec ineracion likelihood and accoun for saic scene componens such as enry/exi locaions ino he scene priors. ACKNOWLEDGMENTS This work was suppored by ONR gran #N REFERENCES [1] R. Gopalan, R. Li, and R. Chellappa, Domain adapaion for objec recogniion: An unsupervised approach, in Compuer Vision (ICCV), 2011 IEEE Inernaional Conference on. IEEE, 2011, pp , 3 [2] F. Fleure, J. Berclaz, R. Lengagne, and P. Fua, Mulicamera People Tracking wih a Probabilisic Occupancy Map, IEEE Transacions on Paern Analysis and Machine Inelligence, vol. 30, no. 2, pp , [3] A. T. Kamal, J. A. Farrell, and A. K. Roy-Chowdhury, Informaion consensus for disribued muli-arge racking, in IEEE Conf. on Compuer Vision and Paern Recogniion, vol. 2, , 6 [4] A. T. Kamal, C. Ding, B. Song, J. A. Farrell, and A. Roy-Chowdhury, A generalized kalman consensus filer for wide-area video neworks, in Decision and Conrol and European Conrol Conference (CDC-ECC), h IEEE Conference on. IEEE, 2011, pp [5] A. T. Kamal, J. A. Farrell, and A. K. Roy-Chowdhury, Informaion weighed consensus, in Decision and Conrol (CDC), 2012 IEEE 51s Annual Conference on. IEEE, 2012, pp , 6 [6] J. Corander, M. Ekdahl, and T. Koski, Parallell ineracing mcmc for learning of opologies of graphical models, Daa mining and knowledge discovery, vol. 17, no. 3, pp , , 5 [7] C. Ding, B. Song, A. Morye, J. Farrell, and A. Roy-Chowdhury, Collaboraive sensing in a disribued pz camera nework, Image Processing, IEEE Transacions on, vol. 21, no. 7, pp , july [8] W. Qu, D. Schonfeld, and M. Mohamed, Disribued Bayesian Muliple-Targe Tracking in Crowded Environmens Using Muliple Collaboraive Cameras, EURASIP Journal on Advances in Signal Processing, vol. 2007, no. 1, pp , Jan [9] Z. Ni, S. Sunderrajan, A. Rahimi, and B. Manjunah, Disribued paricle filer racking wih online muliple insance learning in a camera sensor nework, in Image Processing (ICIP), h IEEE Inernaional Conference on, sep. 2010, pp , 5, 6 [10] I. Saleemi, K. Shafique, and M. Shah, Probabilisic modeling of scene dynamics for applicaions in visual surveillance, Paern Analysis and Machine Inelligence, IEEE Transacions on, vol. 31, no. 8, pp , aug , 5 [11] S. Sunderrajan, S. Karhikeyan, and B. Manjunah, Robus muliple objec racking by deecion wih ineracing markov chain mone carlo, in Image Processing (ICIP), h IEEE Inernaional Conference on. IEEE, sep , 5 [12] S. Avidan, Ensemble racking, in Compuer Vision and Paern Recogniion, CVPR IEEE Compuer Sociey Conference on, vol. 2, june 2005, pp vol. 2. 2, 3 [13] B. Babenko, M.-H. Yang, and S. Belongie, Visual racking wih online muliple insance learning, in IEEE Conference on Compuer Vision and Paern Recogniion, 2009, pp , 6 [14] Z. Ni, S. Sunderrajan, A. Rahimi, and B. Manjunah, Paricle filer racking wih online muliple insance learning, in Inernaional Conference on Paern Recogniion, Aug [15] P. M. Roh, C. Leisner, A. Berger, and H. Bischof, Muliple insance learning from muliple cameras, in Compuer Vision and Paern Recogniion Workshops (CVPRW), 2010 IEEE Compuer Sociey Conference on. IEEE, 2010, pp [16] Z. Khan, T. Balch, and F. Dellaer, Mcmc-based paricle filering for racking a variable number of ineracing arges, Paern Analysis and Machine Inelligence, IEEE Transacions on, vol. 27, no. 11, pp , nov [17] N. Dalal and B. Triggs, Hisograms of oriened gradiens for human deecion, in Compuer Vision and Paern Recogniion, CVPR IEEE Compuer Sociey Conference on, vol. 1, june 2005, pp vol [18] K. A. Gallivan, A. Srivasava, X. Liu, and P. Van Dooren, Efficien algorihms for inferences on grassmann manifolds, in Saisical Signal Processing, 2003 IEEE Workshop on. IEEE, 2003, pp [19] C. Vondrick, A. Khosla, T. Malisiewicz, and A. Torralba, Invering and visualizing feaures for objec deecion, Arxiv preprin cs.cv/ , [20] J. Kwon and K. Lee, Visual racking decomposiion, in Compuer Vision and Paern Recogniion (CVPR), 2010 IEEE Conference on. IEEE, 2010, pp , 5 [21] J. Ferryman and A. Shahrokni, Pes2009: Daase and challenge, in Performance Evaluaion of Tracking and Surveillance (PETS-Winer), 2009 Twelfh IEEE Inernaional Workshop on, dec. 2009, pp [22] H. Grabner and H. Bischof, On-line boosing and vision, in Compuer Vision and Paern Recogniion, 2006 IEEE Compuer Sociey Conference on, vol. 1, june 2006, pp [23] S. Hare, A. Saffari, and P. Torr, Sruck: Srucured oupu racking wih kernels, in Compuer Vision (ICCV), 2011 IEEE Inernaional Conference on, nov. 2011, pp [24] D. Comaniciu, V. Ramesh, and P. Meer, Kernel-based objec racking, Paern Analysis and Machine Inelligence, IEEE Transacions on, vol. 25, no. 5, pp , may [25] Y. Bar-Shalom, F. Daum, and J. Huang, The probabilisic daa associaion filer, Conrol Sysems, IEEE, vol. 29, no. 6, pp , [26] R. Olfai-Saber, Kalman-consensus filer: Opimaliy, sabiliy, and performance, in Decision and Conrol, 2009 held joinly wih he h Chinese Conrol Conference. CDC/CCC Proceedings of he 48h IEEE Conference on. IEEE, 2009, pp

Visual Perception as Bayesian Inference. David J Fleet. University of Toronto

Visual Perception as Bayesian Inference. David J Fleet. University of Toronto Visual Percepion as Bayesian Inference David J Flee Universiy of Torono Basic rules of probabiliy sum rule (for muually exclusive a ): produc rule (condiioning): independence (def n ): Bayes rule: marginalizaion:

More information

Robust Visual Tracking for Multiple Targets

Robust Visual Tracking for Multiple Targets Robus Visual Tracking for Muliple Targes Yizheng Cai, Nando de Freias, and James J. Lile Universiy of Briish Columbia, Vancouver, B.C., Canada, V6T 1Z4 {yizhengc, nando, lile}@cs.ubc.ca Absrac. We address

More information

Learning in Games via Opponent Strategy Estimation and Policy Search

Learning in Games via Opponent Strategy Estimation and Policy Search Learning in Games via Opponen Sraegy Esimaion and Policy Search Yavar Naddaf Deparmen of Compuer Science Universiy of Briish Columbia Vancouver, BC yavar@naddaf.name Nando de Freias (Supervisor) Deparmen

More information

Wheelchair-user Detection Combined with Parts-based Tracking

Wheelchair-user Detection Combined with Parts-based Tracking Wheelchair-user Deecion Combined wih Pars-based Tracking Ukyo Tanikawa 1, Yasuomo Kawanishi 1, Daisuke Deguchi 2,IchiroIde 1, Hiroshi Murase 1 and Ryo Kawai 3 1 Graduae School of Informaion Science, Nagoya

More information

Improved TLD Algorithm for Face Tracking

Improved TLD Algorithm for Face Tracking Absrac Improved TLD Algorihm for Face Tracking Huimin Li a, Chaojing Yu b and Jing Chen c Chongqing Universiy of Poss and Telecommunicaions, Chongqing 400065, China a li.huimin666@163.com, b 15023299065@163.com,

More information

J. Vis. Commun. Image R.

J. Vis. Commun. Image R. J. Vis. Commun. Image R. 20 (2009) 9 27 Conens liss available a ScienceDirec J. Vis. Commun. Image R. journal homepage: www.elsevier.com/locae/jvci Face deecion and racking using a Boosed Adapive Paricle

More information

FACIAL ACTION TRACKING USING PARTICLE FILTERS AND ACTIVE APPEARANCE MODELS. Soumya Hamlaoui & Franck Davoine

FACIAL ACTION TRACKING USING PARTICLE FILTERS AND ACTIVE APPEARANCE MODELS. Soumya Hamlaoui & Franck Davoine FACIAL ACTION TRACKING USING PARTICLE FILTERS AND ACTIVE APPEARANCE MODELS Soumya Hamlaoui & Franck Davoine HEUDIASYC Mixed Research Uni, CNRS / Compiègne Universiy of Technology BP 20529, 60205 Compiègne

More information

Image segmentation. Motivation. Objective. Definitions. A classification of segmentation techniques. Assumptions for thresholding

Image segmentation. Motivation. Objective. Definitions. A classification of segmentation techniques. Assumptions for thresholding Moivaion Image segmenaion Which pixels belong o he same objec in an image/video sequence? (spaial segmenaion) Which frames belong o he same video sho? (emporal segmenaion) Which frames belong o he same

More information

A Fast Stereo-Based Multi-Person Tracking using an Approximated Likelihood Map for Overlapping Silhouette Templates

A Fast Stereo-Based Multi-Person Tracking using an Approximated Likelihood Map for Overlapping Silhouette Templates A Fas Sereo-Based Muli-Person Tracking using an Approximaed Likelihood Map for Overlapping Silhouee Templaes Junji Saake Jun Miura Deparmen of Compuer Science and Engineering Toyohashi Universiy of Technology

More information

Tracking Appearances with Occlusions

Tracking Appearances with Occlusions Tracking ppearances wih Occlusions Ying Wu, Ting Yu, Gang Hua Deparmen of Elecrical & Compuer Engineering Norhwesern Universiy 2145 Sheridan oad, Evanson, IL 60208 {yingwu,ingyu,ganghua}@ece.nwu.edu bsrac

More information

Probabilistic Detection and Tracking of Motion Discontinuities

Probabilistic Detection and Tracking of Motion Discontinuities Probabilisic Deecion and Tracking of Moion Disconinuiies Michael J. Black David J. Flee Xerox Palo Alo Research Cener 3333 Coyoe Hill Road Palo Alo, CA 94304 fblack,fleeg@parc.xerox.com hp://www.parc.xerox.com/fblack,fleeg/

More information

MODEL BASED TECHNIQUE FOR VEHICLE TRACKING IN TRAFFIC VIDEO USING SPATIAL LOCAL FEATURES

MODEL BASED TECHNIQUE FOR VEHICLE TRACKING IN TRAFFIC VIDEO USING SPATIAL LOCAL FEATURES MODEL BASED TECHNIQUE FOR VEHICLE TRACKING IN TRAFFIC VIDEO USING SPATIAL LOCAL FEATURES Arun Kumar H. D. 1 and Prabhakar C. J. 2 1 Deparmen of Compuer Science, Kuvempu Universiy, Shimoga, India ABSTRACT

More information

Video-Based Face Recognition Using Probabilistic Appearance Manifolds

Video-Based Face Recognition Using Probabilistic Appearance Manifolds Video-Based Face Recogniion Using Probabilisic Appearance Manifolds Kuang-Chih Lee Jeffrey Ho Ming-Hsuan Yang David Kriegman klee10@uiuc.edu jho@cs.ucsd.edu myang@honda-ri.com kriegman@cs.ucsd.edu Compuer

More information

Visual Indoor Localization with a Floor-Plan Map

Visual Indoor Localization with a Floor-Plan Map Visual Indoor Localizaion wih a Floor-Plan Map Hang Chu Dep. of ECE Cornell Universiy Ihaca, NY 14850 hc772@cornell.edu Absrac In his repor, a indoor localizaion mehod is presened. The mehod akes firsperson

More information

A Hierarchical Object Recognition System Based on Multi-scale Principal Curvature Regions

A Hierarchical Object Recognition System Based on Multi-scale Principal Curvature Regions A Hierarchical Objec Recogniion Sysem Based on Muli-scale Principal Curvaure Regions Wei Zhang, Hongli Deng, Thomas G Dieerich and Eric N Morensen School of Elecrical Engineering and Compuer Science Oregon

More information

Upper Body Tracking for Human-Machine Interaction with a Moving Camera

Upper Body Tracking for Human-Machine Interaction with a Moving Camera The 2009 IEEE/RSJ Inernaional Conference on Inelligen Robos and Sysems Ocober -5, 2009 S. Louis, USA Upper Body Tracking for Human-Machine Ineracion wih a Moving Camera Yi-Ru Chen, Cheng-Ming Huang, and

More information

Robot localization under perceptual aliasing conditions based on laser reflectivity using particle filter

Robot localization under perceptual aliasing conditions based on laser reflectivity using particle filter Robo localizaion under percepual aliasing condiions based on laser refleciviy using paricle filer DongXiang Zhang, Ryo Kurazume, Yumi Iwashia, Tsuomu Hasegawa Absrac Global localizaion, which deermines

More information

IntentSearch:Capturing User Intention for One-Click Internet Image Search

IntentSearch:Capturing User Intention for One-Click Internet Image Search JOURNAL OF L A T E X CLASS FILES, VOL. 6, NO. 1, JANUARY 2010 1 InenSearch:Capuring User Inenion for One-Click Inerne Image Search Xiaoou Tang, Fellow, IEEE, Ke Liu, Jingyu Cui, Suden Member, IEEE, Fang

More information

Robust Multi-view Face Detection Using Error Correcting Output Codes

Robust Multi-view Face Detection Using Error Correcting Output Codes Robus Muli-view Face Deecion Using Error Correcing Oupu Codes Hongming Zhang,2, Wen GaoP P, Xilin Chen 2, Shiguang Shan 2, and Debin Zhao Deparmen of Compuer Science and Engineering, Harbin Insiue of Technolog

More information

CAMERA CALIBRATION BY REGISTRATION STEREO RECONSTRUCTION TO 3D MODEL

CAMERA CALIBRATION BY REGISTRATION STEREO RECONSTRUCTION TO 3D MODEL CAMERA CALIBRATION BY REGISTRATION STEREO RECONSTRUCTION TO 3D MODEL Klečka Jan Docoral Degree Programme (1), FEEC BUT E-mail: xkleck01@sud.feec.vubr.cz Supervised by: Horák Karel E-mail: horak@feec.vubr.cz

More information

Implementing Ray Casting in Tetrahedral Meshes with Programmable Graphics Hardware (Technical Report)

Implementing Ray Casting in Tetrahedral Meshes with Programmable Graphics Hardware (Technical Report) Implemening Ray Casing in Terahedral Meshes wih Programmable Graphics Hardware (Technical Repor) Marin Kraus, Thomas Erl March 28, 2002 1 Inroducion Alhough cell-projecion, e.g., [3, 2], and resampling,

More information

Rao-Blackwellized Particle Filtering for Probing-Based 6-DOF Localization in Robotic Assembly

Rao-Blackwellized Particle Filtering for Probing-Based 6-DOF Localization in Robotic Assembly MITSUBISHI ELECTRIC RESEARCH LABORATORIES hp://www.merl.com Rao-Blackwellized Paricle Filering for Probing-Based 6-DOF Localizaion in Roboic Assembly Yuichi Taguchi, Tim Marks, Haruhisa Okuda TR1-8 June

More information

A Matching Algorithm for Content-Based Image Retrieval

A Matching Algorithm for Content-Based Image Retrieval A Maching Algorihm for Conen-Based Image Rerieval Sue J. Cho Deparmen of Compuer Science Seoul Naional Universiy Seoul, Korea Absrac Conen-based image rerieval sysem rerieves an image from a daabase using

More information

Evaluation and Improvement of Region-based Motion Segmentation

Evaluation and Improvement of Region-based Motion Segmentation Evaluaion and Improvemen of Region-based Moion Segmenaion Mark Ross Universiy Koblenz-Landau, Insiue of Compuaional Visualisics, Universiässraße 1, 56070 Koblenz, Germany Email: ross@uni-koblenz.de Absrac

More information

Occlusion-Free Hand Motion Tracking by Multiple Cameras and Particle Filtering with Prediction

Occlusion-Free Hand Motion Tracking by Multiple Cameras and Particle Filtering with Prediction 58 IJCSNS Inernaional Journal of Compuer Science and Nework Securiy, VOL.6 No.10, Ocober 006 Occlusion-Free Hand Moion Tracking by Muliple Cameras and Paricle Filering wih Predicion Makoo Kao, and Gang

More information

A Novel Approach for Monocular 3D Object Tracking in Cluttered Environment

A Novel Approach for Monocular 3D Object Tracking in Cluttered Environment Inernaional Journal of Compuaional Inelligence Research ISSN 0973-1873 Volume 13, Number 5 (2017), pp. 851-864 Research India Publicaions hp://www.ripublicaion.com A Novel Approach for Monocular 3D Objec

More information

Deformable Parts Correlation Filters for Robust Visual Tracking

Deformable Parts Correlation Filters for Robust Visual Tracking PAPER UNDER REVISION Deformable Pars Correlaion Filers for Robus Visual Tracking Alan Lukežič, Luka Čehovin, Member, IEEE, and Maej Krisan, Member, IEEE ha par-based models should be considered in a layered

More information

IROS 2015 Workshop on On-line decision-making in multi-robot coordination (DEMUR 15)

IROS 2015 Workshop on On-line decision-making in multi-robot coordination (DEMUR 15) IROS 2015 Workshop on On-line decision-making in muli-robo coordinaion () OPTIMIZATION-BASED COOPERATIVE MULTI-ROBOT TARGET TRACKING WITH REASONING ABOUT OCCLUSIONS KAROL HAUSMAN a,, GREGORY KAHN b, SACHIN

More information

Multi-camera multi-object voxel-based Monte Carlo 3D tracking strategies

Multi-camera multi-object voxel-based Monte Carlo 3D tracking strategies RESEARCH Open Access Muli-camera muli-objec voxel-based Mone Carlo 3D racking sraegies Crisian Canon-Ferrer *, Josep R Casas, Monse Pardàs and Enric Mone Absrac This aricle presens a new approach o he

More information

Real Time Integral-Based Structural Health Monitoring

Real Time Integral-Based Structural Health Monitoring Real Time Inegral-Based Srucural Healh Monioring The nd Inernaional Conference on Sensing Technology ICST 7 J. G. Chase, I. Singh-Leve, C. E. Hann, X. Chen Deparmen of Mechanical Engineering, Universiy

More information

Detection and segmentation of moving objects in highly dynamic scenes

Detection and segmentation of moving objects in highly dynamic scenes Deecion and segmenaion of moving objecs in highly dynamic scenes Aurélie Bugeau Parick Pérez INRIA, Cenre Rennes - Breagne Alanique Universié de Rennes, Campus de Beaulieu, 35 042 Rennes Cedex, France

More information

STEREO PLANE MATCHING TECHNIQUE

STEREO PLANE MATCHING TECHNIQUE STEREO PLANE MATCHING TECHNIQUE Commission III KEY WORDS: Sereo Maching, Surface Modeling, Projecive Transformaion, Homography ABSTRACT: This paper presens a new ype of sereo maching algorihm called Sereo

More information

Video Content Description Using Fuzzy Spatio-Temporal Relations

Video Content Description Using Fuzzy Spatio-Temporal Relations Proceedings of he 4s Hawaii Inernaional Conference on Sysem Sciences - 008 Video Conen Descripion Using Fuzzy Spaio-Temporal Relaions rchana M. Rajurkar *, R.C. Joshi and Sananu Chaudhary 3 Dep of Compuer

More information

Multi-Target Detection and Tracking from a Single Camera in Unmanned Aerial Vehicles (UAVs)

Multi-Target Detection and Tracking from a Single Camera in Unmanned Aerial Vehicles (UAVs) 2016 IEEE/RSJ Inernaional Conference on Inelligen Robos and Sysems (IROS) Daejeon Convenion Cener Ocober 9-14, 2016, Daejeon, Korea Muli-Targe Deecion and Tracking from a Single Camera in Unmanned Aerial

More information

Real-Time Non-Rigid Multi-Frame Depth Video Super-Resolution

Real-Time Non-Rigid Multi-Frame Depth Video Super-Resolution Real-Time Non-Rigid Muli-Frame Deph Video Super-Resoluion Kassem Al Ismaeil 1, Djamila Aouada 1, Thomas Solignac 2, Bruno Mirbach 2, Björn Oersen 1 1 Inerdisciplinary Cenre for Securiy, Reliabiliy, and

More information

We are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists. International authors and editors

We are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists. International authors and editors We are InechOpen, he world s leading publisher of Open Access books Buil by scieniss, for scieniss 4,000 116,000 120M Open access books available Inernaional auhors and ediors Downloads Our auhors are

More information

Viewpoint Invariant 3D Landmark Model Inference from Monocular 2D Images Using Higher-Order Priors

Viewpoint Invariant 3D Landmark Model Inference from Monocular 2D Images Using Higher-Order Priors Viewpoin Invarian 3D Landmark Model Inference from Monocular 2D Images Using Higher-Order Priors Chaohui Wang 1,2, Yun Zeng 3, Loic Simon 1, Ioannis Kakadiaris 4, Dimiris Samaras 3, Nikos Paragios 1,2

More information

Analysis of Various Types of Bugs in the Object Oriented Java Script Language Coding

Analysis of Various Types of Bugs in the Object Oriented Java Script Language Coding Indian Journal of Science and Technology, Vol 8(21), DOI: 10.17485/ijs/2015/v8i21/69958, Sepember 2015 ISSN (Prin) : 0974-6846 ISSN (Online) : 0974-5645 Analysis of Various Types of Bugs in he Objec Oriened

More information

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS PART A: SYSTEMS AND HUMANS 1

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS PART A: SYSTEMS AND HUMANS 1 TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS PART A: SYSTEMS AND HUMANS 1 Adapive Appearance Model and Condensaion Algorihm for Robus Face Tracking Yui Man Lui, Suden Member,, J. Ross Beveridge, Member,,

More information

Real-time 2D Video/3D LiDAR Registration

Real-time 2D Video/3D LiDAR Registration Real-ime 2D Video/3D LiDAR Regisraion C. Bodenseiner Fraunhofer IOSB chrisoph.bodenseiner@iosb.fraunhofer.de M. Arens Fraunhofer IOSB michael.arens@iosb.fraunhofer.de Absrac Progress in LiDAR scanning

More information

Improving Occupancy Grid FastSLAM by Integrating Navigation Sensors

Improving Occupancy Grid FastSLAM by Integrating Navigation Sensors Improving Occupancy Grid FasSLAM by Inegraing Navigaion Sensors Chrisopher Weyers Sensors Direcorae Air Force Research Laboraory Wrigh-Paerson AFB, OH 45433 Gilber Peerson Deparmen of Elecrical and Compuer

More information

Weighted Voting in 3D Random Forest Segmentation

Weighted Voting in 3D Random Forest Segmentation Weighed Voing in 3D Random Fores Segmenaion M. Yaqub,, P. Mahon 3, M. K. Javaid, C. Cooper, J. A. Noble NDORMS, Universiy of Oxford, IBME, Deparmen of Engineering Science, Universiy of Oxford, 3 MRC Epidemiology

More information

MORPHOLOGICAL SEGMENTATION OF IMAGE SEQUENCES

MORPHOLOGICAL SEGMENTATION OF IMAGE SEQUENCES MORPHOLOGICAL SEGMENTATION OF IMAGE SEQUENCES B. MARCOTEGUI and F. MEYER Ecole des Mines de Paris, Cenre de Morphologie Mahémaique, 35, rue Sain-Honoré, F 77305 Fonainebleau Cedex, France Absrac. In image

More information

Simultaneous Localization and Mapping with Stereo Vision

Simultaneous Localization and Mapping with Stereo Vision Simulaneous Localizaion and Mapping wih Sereo Vision Mahew N. Dailey Compuer Science and Informaion Managemen Asian Insiue of Technology Pahumhani, Thailand Email: mdailey@ai.ac.h Manukid Parnichkun Mecharonics

More information

Joint Feature Learning With Robust Local Ternary Pattern for Face Recognition

Joint Feature Learning With Robust Local Ternary Pattern for Face Recognition Join Feaure Learning Wih Robus Local Ternary Paern for Face Recogniion Yuvaraju.M 1, Shalini.S 1 Assisan Professor, Deparmen of Elecrical and Elecronics Engineering, Anna Universiy Regional Campus, Coimbaore,

More information

COSC 3213: Computer Networks I Chapter 6 Handout # 7

COSC 3213: Computer Networks I Chapter 6 Handout # 7 COSC 3213: Compuer Neworks I Chaper 6 Handou # 7 Insrucor: Dr. Marvin Mandelbaum Deparmen of Compuer Science York Universiy F05 Secion A Medium Access Conrol (MAC) Topics: 1. Muliple Access Communicaions:

More information

Mobile Robots Mapping

Mobile Robots Mapping Mobile Robos Mapping 1 Roboics is Easy conrol behavior percepion modelling domain model environmen model informaion exracion raw daa planning ask cogniion reasoning pah planning navigaion pah execuion

More information

Gender Classification of Faces Using Adaboost*

Gender Classification of Faces Using Adaboost* Gender Classificaion of Faces Using Adaboos* Rodrigo Verschae 1,2,3, Javier Ruiz-del-Solar 1,2, and Mauricio Correa 1,2 1 Deparmen of Elecrical Engineering, Universidad de Chile 2 Cener for Web Research,

More information

Moving Object Detection Using MRF Model and Entropy based Adaptive Thresholding

Moving Object Detection Using MRF Model and Entropy based Adaptive Thresholding Moving Objec Deecion Using MRF Model and Enropy based Adapive Thresholding Badri Narayan Subudhi, Pradipa Kumar Nanda and Ashish Ghosh Machine Inelligence Uni, Indian Saisical Insiue, Kolkaa, 700108, India,

More information

Track-based and object-based occlusion for people tracking refinement in indoor surveillance

Track-based and object-based occlusion for people tracking refinement in indoor surveillance Trac-based and objec-based occlusion for people racing refinemen in indoor surveillance R. Cucchiara, C. Grana, G. Tardini Diparimeno di Ingegneria Informaica - Universiy of Modena and Reggio Emilia Via

More information

Robust 3D Visual Tracking Using Particle Filtering on the SE(3) Group

Robust 3D Visual Tracking Using Particle Filtering on the SE(3) Group Robus 3D Visual Tracking Using Paricle Filering on he SE(3) Group Changhyun Choi and Henrik I. Chrisensen Roboics & Inelligen Machines, College of Compuing Georgia Insiue of Technology Alana, GA 3332,

More information

A Face Detection Method Based on Skin Color Model

A Face Detection Method Based on Skin Color Model A Face Deecion Mehod Based on Skin Color Model Dazhi Zhang Boying Wu Jiebao Sun Qinglei Liao Deparmen of Mahemaics Harbin Insiue of Technology Harbin China 150000 Zhang_dz@163.com mahwby@hi.edu.cn sunjiebao@om.com

More information

Detection of salient objects with focused attention based on spatial and temporal coherence

Detection of salient objects with focused attention based on spatial and temporal coherence ricle Informaion Processing Technology pril 2011 Vol.56 No.10: 1055 1062 doi: 10.1007/s11434-010-4387-1 SPECIL TOPICS: Deecion of salien objecs wih focused aenion based on spaial and emporal coherence

More information

Improving the Efficiency of Dynamic Service Provisioning in Transport Networks with Scheduled Services

Improving the Efficiency of Dynamic Service Provisioning in Transport Networks with Scheduled Services Improving he Efficiency of Dynamic Service Provisioning in Transpor Neworks wih Scheduled Services Ralf Hülsermann, Monika Jäger and Andreas Gladisch Technologiezenrum, T-Sysems, Goslarer Ufer 35, D-1585

More information

MOTION TRACKING is a fundamental capability that

MOTION TRACKING is a fundamental capability that TECHNICAL REPORT CRES-05-008, CENTER FOR ROBOTICS AND EMBEDDED SYSTEMS, UNIVERSITY OF SOUTHERN CALIFORNIA 1 Real-ime Moion Tracking from a Mobile Robo Boyoon Jung, Suden Member, IEEE, Gaurav S. Sukhame,

More information

Learning nonlinear appearance manifolds for robot localization

Learning nonlinear appearance manifolds for robot localization Learning nonlinear appearance manifolds for robo localizaion Jihun Hamm, Yuanqing Lin, and Daniel. D. Lee GRAS Lab, Deparmen of Elecrical and Sysems Engineering Universiy of ennsylvania, hiladelphia, A

More information

Why not experiment with the system itself? Ways to study a system System. Application areas. Different kinds of systems

Why not experiment with the system itself? Ways to study a system System. Application areas. Different kinds of systems Simulaion Wha is simulaion? Simple synonym: imiaion We are ineresed in sudying a Insead of experimening wih he iself we experimen wih a model of he Experimen wih he Acual Ways o sudy a Sysem Experimen

More information

Hidden Markov Model and Chapman Kolmogrov for Protein Structures Prediction from Images

Hidden Markov Model and Chapman Kolmogrov for Protein Structures Prediction from Images Hidden Markov Model and Chapman Kolmogrov for Proein Srucures Predicion from Images Md.Sarwar Kamal 1, Linkon Chowdhury 2, Mohammad Ibrahim Khan 2, Amira S. Ashour 3, João Manuel R.S. Tavares 4, Nilanjan

More information

Robust LSTM-Autoencoders for Face De-Occlusion in the Wild

Robust LSTM-Autoencoders for Face De-Occlusion in the Wild IEEE TRANSACTIONS ON IMAGE PROCESSING, DRAFT 1 Robus LSTM-Auoencoders for Face De-Occlusion in he Wild Fang Zhao, Jiashi Feng, Jian Zhao, Wenhan Yang, Shuicheng Yan arxiv:1612.08534v1 [cs.cv] 27 Dec 2016

More information

Detection Tracking and Recognition of Human Poses for a Real Time Spatial Game

Detection Tracking and Recognition of Human Poses for a Real Time Spatial Game Deecion Tracking and Recogniion of Human Poses for a Real Time Spaial Game Feifei Huo, Emile A. Hendriks, A.H.J. Oomes Delf Universiy of Technology The Neherlands f.huo@udelf.nl Pascal van Beek, Remco

More information

In Proceedings of CVPR '96. Structure and Motion of Curved 3D Objects from. using these methods [12].

In Proceedings of CVPR '96. Structure and Motion of Curved 3D Objects from. using these methods [12]. In Proceedings of CVPR '96 Srucure and Moion of Curved 3D Objecs from Monocular Silhouees B Vijayakumar David J Kriegman Dep of Elecrical Engineering Yale Universiy New Haven, CT 652-8267 Jean Ponce Compuer

More information

A GRAPHICS PROCESSING UNIT IMPLEMENTATION OF THE PARTICLE FILTER

A GRAPHICS PROCESSING UNIT IMPLEMENTATION OF THE PARTICLE FILTER A GRAPHICS PROCESSING UNIT IMPLEMENTATION OF THE PARTICLE FILTER ABSTRACT Modern graphics cards for compuers, and especially heir graphics processing unis (GPUs), are designed for fas rendering of graphics.

More information

Real time 3D face and facial feature tracking

Real time 3D face and facial feature tracking J Real-Time Image Proc (2007) 2:35 44 DOI 10.1007/s11554-007-0032-2 ORIGINAL RESEARCH PAPER Real ime 3D face and facial feaure racking Fadi Dornaika Æ Javier Orozco Received: 23 November 2006 / Acceped:

More information

Simple Network Management Based on PHP and SNMP

Simple Network Management Based on PHP and SNMP Simple Nework Managemen Based on PHP and SNMP Krasimir Trichkov, Elisavea Trichkova bsrac: This paper aims o presen simple mehod for nework managemen based on SNMP - managemen of Cisco rouer. The paper

More information

An Iterative Scheme for Motion-Based Scene Segmentation

An Iterative Scheme for Motion-Based Scene Segmentation An Ieraive Scheme for Moion-Based Scene Segmenaion Alexander Bachmann and Hildegard Kuehne Deparmen for Measuremen and Conrol Insiue for Anhropomaics Universiy of Karlsruhe (H), 76 131 Karlsruhe, Germany

More information

Design Alternatives for a Thin Lens Spatial Integrator Array

Design Alternatives for a Thin Lens Spatial Integrator Array Egyp. J. Solids, Vol. (7), No. (), (004) 75 Design Alernaives for a Thin Lens Spaial Inegraor Array Hala Kamal *, Daniel V azquez and Javier Alda and E. Bernabeu Opics Deparmen. Universiy Compluense of

More information

ACQUIRING high-quality and well-defined depth data. Online Temporally Consistent Indoor Depth Video Enhancement via Static Structure

ACQUIRING high-quality and well-defined depth data. Online Temporally Consistent Indoor Depth Video Enhancement via Static Structure SUBMITTED TO TRANSACTION ON IMAGE PROCESSING 1 Online Temporally Consisen Indoor Deph Video Enhancemen via Saic Srucure Lu Sheng, Suden Member, IEEE, King Ngi Ngan, Fellow, IEEE, Chern-Loon Lim and Songnan

More information

EECS 487: Interactive Computer Graphics

EECS 487: Interactive Computer Graphics EECS 487: Ineracive Compuer Graphics Lecure 7: B-splines curves Raional Bézier and NURBS Cubic Splines A represenaion of cubic spline consiss of: four conrol poins (why four?) hese are compleely user specified

More information

SENSING using 3D technologies, structured light cameras

SENSING using 3D technologies, structured light cameras IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 39, NO. 10, OCTOBER 2017 2045 Real-Time Enhancemen of Dynamic Deph Videos wih Non-Rigid Deformaions Kassem Al Ismaeil, Suden Member,

More information

Tracking a Large Number of Objects from Multiple Views

Tracking a Large Number of Objects from Multiple Views Tracking a Large Number of Objecs from Muliple Views Zheng Wu 1, Nickolay I. Hrisov 2, Tyson L. Hedrick 3, Thomas H. Kun 2, Margri Beke 1 1 Deparmen of Compuer Science, Boson Universiy 2 Deparmen of Biology,

More information

Reinforcement Learning by Policy Improvement. Making Use of Experiences of The Other Tasks. Hajime Kimura and Shigenobu Kobayashi

Reinforcement Learning by Policy Improvement. Making Use of Experiences of The Other Tasks. Hajime Kimura and Shigenobu Kobayashi Reinforcemen Learning by Policy Improvemen Making Use of Experiences of The Oher Tasks Hajime Kimura and Shigenobu Kobayashi Tokyo Insiue of Technology, JAPAN genfe.dis.iech.ac.jp, kobayasidis.iech.ac.jp

More information

Multi-View 3D Human Tracking in Crowded Scenes

Multi-View 3D Human Tracking in Crowded Scenes Proceedings of he Thirieh AAAI Conference on Arificial Inelligence (AAAI-16) Muli-View 3D Human Tracking in Crowded Scenes Xiaobai Liu Deparmen of Compuer Science, San Diego Sae Universiy GMCS Building,

More information

Nonparametric CUSUM Charts for Process Variability

Nonparametric CUSUM Charts for Process Variability Journal of Academia and Indusrial Research (JAIR) Volume 3, Issue June 4 53 REEARCH ARTICLE IN: 78-53 Nonparameric CUUM Chars for Process Variabiliy D.M. Zombade and V.B. Ghue * Dep. of aisics, Walchand

More information

Multi-Scale Object Candidates for Generic Object Tracking in Street Scenes

Multi-Scale Object Candidates for Generic Object Tracking in Street Scenes Muli-Scale Objec Candidaes for Generic Objec Tracking in Sree Scenes Aljoša Ošep, Alexander Hermans, Francis Engelmann, Dirk Klosermann, Markus Mahias and Basian Leibe Absrac Mos vision based sysems for

More information

A Framework for Applying Point Clouds Grabbed by Multi-Beam LIDAR in Perceiving the Driving Environment

A Framework for Applying Point Clouds Grabbed by Multi-Beam LIDAR in Perceiving the Driving Environment Sensors 215, 15, 21931-21956; doi:1.339/s15921931 Aricle OPEN ACCESS sensors ISSN 1424-822 www.mdpi.com/journal/sensors A Framewor for Applying Poin Clouds Grabbed by Muli-Beam LIDAR in Perceiving he Driving

More information

A Bayesian Approach to Video Object Segmentation via Merging 3D Watershed Volumes

A Bayesian Approach to Video Object Segmentation via Merging 3D Watershed Volumes A Bayesian Approach o Video Objec Segmenaion via Merging 3D Waershed Volumes Yu-Pao Tsai 1,3, Chih-Chuan Lai 1,2, Yi-Ping Hung 1,2, and Zen-Chung Shih 3 1 Insiue of Informaion Science, Academia Sinica,

More information

Tracking a Large Number of Objects from Multiple Views

Tracking a Large Number of Objects from Multiple Views Boson Universiy Compuer Science Deparmen Technical Repor BUCS-TR 2009-005 Tracking a Large Number of Objecs from Muliple Views Zheng Wu 1, Nickolay I. Hrisov 2, Tyson L. Hedrick 3, Thomas H. Kun 2, Margri

More information

TrackNet: Simultaneous Detection and Tracking of Multiple Objects

TrackNet: Simultaneous Detection and Tracking of Multiple Objects TrackNe: Simulaneous Deecion and Tracking of Muliple Objecs Chenge Li New York Universiy cl2840@nyu.edu Gregory Dobler New York Universiy greg.dobler@nyu.edu Yilin Song New York Universiy ys1297@nyu.edu

More information

A new algorithm for small object tracking based on super-resolution technique

A new algorithm for small object tracking based on super-resolution technique A new algorihm for small objec racking based on super-resoluion echnique Yabunayya Habibi, Dwi Rana Sulisyaningrum, and Budi Seiyono Ciaion: AIP Conference Proceedings 1867, 020024 (2017); doi: 10.1063/1.4994427

More information

Robust Segmentation and Tracking of Colored Objects in Video

Robust Segmentation and Tracking of Colored Objects in Video IEEE TRANSACTIONS ON CSVT, VOL. 4, NO. 6, 2004 Robus Segmenaion and Tracking of Colored Objecs in Video Theo Gevers, member, IEEE Absrac Segmening and racking of objecs in video is of grea imporance for

More information

Sequential Monte Carlo Tracking for Marginal Artery Segmentation on CT Angiography by Multiple Cue Fusion

Sequential Monte Carlo Tracking for Marginal Artery Segmentation on CT Angiography by Multiple Cue Fusion Sequenial Mone Carlo Tracking for Marginal Arery Segmenaion on CT Angiography by Muliple Cue Fusion Shijun Wang, Brandon Peplinski, Le Lu, Weidong Zhang, Jianfei Liu, Zhuoshi Wei, and Ronald M. Summers

More information

The Impact of Product Development on the Lifecycle of Defects

The Impact of Product Development on the Lifecycle of Defects The Impac of Produc Developmen on he Lifecycle of Rudolf Ramler Sofware Compeence Cener Hagenberg Sofware Park 21 A-4232 Hagenberg, Ausria +43 7236 3343 872 rudolf.ramler@scch.a ABSTRACT This paper invesigaes

More information

Track and Cut: simultaneous tracking and segmentation of multiple objects with graph cuts

Track and Cut: simultaneous tracking and segmentation of multiple objects with graph cuts INSTITUT NATIONAL DE RECHERCHE EN INFORMATIQUE ET EN AUTOMATIQUE Track and Cu: simulaneous racking and segmenaion of muliple objecs wih graph cus Aurélie Bugeau Parick Pérez N 6337 Ocober 2007 Thèmes COM

More information

Michiel Helder and Marielle C.T.A Geurts. Hoofdkantoor PTT Post / Dutch Postal Services Headquarters

Michiel Helder and Marielle C.T.A Geurts. Hoofdkantoor PTT Post / Dutch Postal Services Headquarters SHORT TERM PREDICTIONS A MONITORING SYSTEM by Michiel Helder and Marielle C.T.A Geurs Hoofdkanoor PTT Pos / Duch Posal Services Headquarers Keywords macro ime series shor erm predicions ARIMA-models faciliy

More information

Modeling and Tracking of Dynamic Obstacles for Logistic Plants using Omnidirectional Stereo Vision

Modeling and Tracking of Dynamic Obstacles for Logistic Plants using Omnidirectional Stereo Vision Modeling and Tracing of Dynamic Obsacles for Logisic Plans using Omnidirecional Sereo Vision Andrei Vaavu, Arhur D. Cosea, and Sergiu Nedevschi, Members, IEEE Absrac In his wor we presen an obsacle deecion

More information

Motion Level-of-Detail: A Simplification Method on Crowd Scene

Motion Level-of-Detail: A Simplification Method on Crowd Scene Moion Level-of-Deail: A Simplificaion Mehod on Crowd Scene Absrac Junghyun Ahn VR lab, EECS, KAIST ChocChoggi@vr.kais.ac.kr hp://vr.kais.ac.kr/~zhaoyue Recen echnological improvemen in characer animaion

More information

Parallel and Distributed Systems for Constructive Neural Network Learning*

Parallel and Distributed Systems for Constructive Neural Network Learning* Parallel and Disribued Sysems for Consrucive Neural Nework Learning* J. Flecher Z. Obradovi School of Elecrical Engineering and Compuer Science Washingon Sae Universiy Pullman WA 99164-2752 Absrac A consrucive

More information

An Adaptive Spatial Depth Filter for 3D Rendering IP

An Adaptive Spatial Depth Filter for 3D Rendering IP JOURNAL OF SEMICONDUCTOR TECHNOLOGY AND SCIENCE, VOL.3, NO. 4, DECEMBER, 23 175 An Adapive Spaial Deph Filer for 3D Rendering IP Chang-Hyo Yu and Lee-Sup Kim Absrac In his paper, we presen a new mehod

More information

Coded Caching with Multiple File Requests

Coded Caching with Multiple File Requests Coded Caching wih Muliple File Requess Yi-Peng Wei Sennur Ulukus Deparmen of Elecrical and Compuer Engineering Universiy of Maryland College Park, MD 20742 ypwei@umd.edu ulukus@umd.edu Absrac We sudy a

More information

Difficulty-aware Hybrid Search in Peer-to-Peer Networks

Difficulty-aware Hybrid Search in Peer-to-Peer Networks Difficuly-aware Hybrid Search in Peer-o-Peer Neworks Hanhua Chen, Hai Jin, Yunhao Liu, Lionel M. Ni School of Compuer Science and Technology Huazhong Univ. of Science and Technology {chenhanhua, hjin}@hus.edu.cn

More information

THE goal of this work is to develop statistical models for

THE goal of this work is to develop statistical models for IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 32, NO. 4, APRIL 2010 579 Nonsaionary Shape Aciviies: Dynamic Models for Landmark Shape Change and Applicaions Samarji Das, Suden Member,

More information

Traditional Rendering (Ray Tracing and Radiosity)

Traditional Rendering (Ray Tracing and Radiosity) Tradiional Rendering (Ray Tracing and Radiosiy) CS 517 Fall 2002 Compuer Science Cornell Universiy Bidirecional Reflecance (BRDF) λ direcional diffuse specular θ uniform diffuse τ σ BRDF Bidirecional Reflecance

More information

An Improved Square-Root Nyquist Shaping Filter

An Improved Square-Root Nyquist Shaping Filter An Improved Square-Roo Nyquis Shaping Filer fred harris San Diego Sae Universiy fred.harris@sdsu.edu Sridhar Seshagiri San Diego Sae Universiy Seshigar.@engineering.sdsu.edu Chris Dick Xilinx Corp. chris.dick@xilinx.com

More information

MoBAN: A Configurable Mobility Model for Wireless Body Area Networks

MoBAN: A Configurable Mobility Model for Wireless Body Area Networks MoBAN: A Configurable Mobiliy Model for Wireless Body Area Neworks Majid Nabi 1, Marc Geilen 1, Twan Basen 1,2 1 Deparmen of Elecrical Engineering, Eindhoven Universiy of Technology, he Neherlands 2 Embedded

More information

Stereoscopic Neural Style Transfer

Stereoscopic Neural Style Transfer Sereoscopic Neural Syle Transfer Dongdong Chen 1 Lu Yuan 2, Jing Liao 2, Nenghai Yu 1, Gang Hua 2 1 Universiy of Science and Technology of China 2 Microsof Research cd722522@mail.usc.edu.cn, {luyuan,jliao}@microsof.com,

More information

Improving Ranking of Search Engines Results Based on Power Links

Improving Ranking of Search Engines Results Based on Power Links IPASJ Inernaional Journal of Informaion Technology (IIJIT) Web Sie: hp://www.ipasj.org/iijit/iijit.hm A Publisher for Research Moivaion... Email: edioriiji@ipasj.org Volume 2, Issue 9, Sepember 2014 ISSN

More information

Scale Recovery for Monocular Visual Odometry Using Depth Estimated with Deep Convolutional Neural Fields

Scale Recovery for Monocular Visual Odometry Using Depth Estimated with Deep Convolutional Neural Fields Scale Recovery for Monocular Visual Odomery Using Deph Esimaed wih Deep Convoluional Neural Fields Xiaochuan Yin, Xiangwei Wang, Xiaoguo Du, Qijun Chen Tongji Universiy yinxiaochuan@homail.com,wangxiangwei.cpp@gmail.com,

More information

A GRAPHICS PROCESSING UNIT IMPLEMENTATION OF THE PARTICLE FILTER

A GRAPHICS PROCESSING UNIT IMPLEMENTATION OF THE PARTICLE FILTER A GRAPHICS PROCESSING UNIT IMPLEMENTATION OF THE PARTICLE FILTER Gusaf Hendeby, Jeroen D. Hol, Rickard Karlsson, Fredrik Gusafsson Deparmen of Elecrical Engineering Auomaic Conrol Linköping Universiy,

More information

A METHOD OF MODELING DEFORMATION OF AN OBJECT EMPLOYING SURROUNDING VIDEO CAMERAS

A METHOD OF MODELING DEFORMATION OF AN OBJECT EMPLOYING SURROUNDING VIDEO CAMERAS A METHOD OF MODELING DEFORMATION OF AN OBJECT EMLOYING SURROUNDING IDEO CAMERAS Joo Kooi TAN, Seiji ISHIKAWA Deparmen of Mechanical and Conrol Engineering Kushu Insiue of Technolog, Japan ehelan@is.cnl.kuech.ac.jp,

More information

AUTOMATIC 3D FACE REGISTRATION WITHOUT INITIALIZATION

AUTOMATIC 3D FACE REGISTRATION WITHOUT INITIALIZATION Chaper 3 AUTOMATIC 3D FACE REGISTRATION WITHOUT INITIALIZATION A. Koschan, V. R. Ayyagari, F. Boughorbel, and M. A. Abidi Imaging, Roboics, and Inelligen Sysems Laboraory, The Universiy of Tennessee, 334

More information