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

Size: px
Start display at page:

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

Transcription

1 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, USA Absrac In his paper, we presen a 3D model-based objec racking approach using edge and keypoin feaures in a paricle filering framework. Edge poins provide 1D informaion for pose esimaion and i is naural o consider muliple hypoheses. Recenly, paricle filering based approaches have been proposed o inegrae muliple hypoheses and have shown good performance, bu mos of he work has made an assumpion ha an iniial pose is given. To remove his assumpion, we employ keypoin feaures for iniializaion of he filer. Given 2D-3D keypoin correspondences, we choose a se of minimum correspondences o calculae a se of possible pose hypoheses. Based on he inlier raio of correspondences, he se of poses are drawn o iniialize paricles. For beer performance, we employ an auoregressive sae dynamics and apply i o a coordinae-invarian paricle filer on he SE(3) group. Based on he number of effecive paricles calculaed during racking, he proposed sysem re-iniializes paricles when he racked objec goes ou of sigh or is occluded. The robusness and accuracy of our approach is demonsraed via comparaive experimens. I. INTRODUCTION From roboic manipulaion o augmened realiy, esimaing poses of objecs is a key ask. Since Harris [1] proposed his early sysem which racks an objec by projecing a 3D CAD model ino a 2D image and aligning he projeced model edges o image edges, here have been acive effors o enhance he early edge-based racking sysem [2], [3]. Edges are employed because hey are easy o compue and invarian o illuminaion and pose changes. However, a criical disadvanage of using edges in visual racking is ha hey look similar o each oher. In general, edge correspondences are deermined by local search based on a prior pose esimae. So he racking performance of an edge-based racker direcly depends on correc pose priors. To improve pose priors, here have been aemps o enhance he pose accuracy beween frames by incorporaing ineres poins [4], [5], [6] or employing addiional sensors [7]. Since ineres poins and heir rich descripors [8], [9] can be exraced and mached well under illuminaion, scale, roaion changes, and reasonable projecion ransformaion, keypoin feaures complemen edges well. Considering muliple edge correspondences was anoher ineres in edge-based racking. Since edges are ambiguous and false edge correspondences direcly lead he racker o false pose esimaes, some approaches have considered muliple edge correspondences [5], [1]. However, heir work was sill limied because only one or wo hypoheses were mainained from he muliple correspondences during racking. Muliple hypoheses racking has been implemened using a paricle filering framework. Isard and Blake [11] applied a paricle filer o 2D visual edge racking and have shown grea poenial. Affine 2D visual rackers have also been proposed in a paricle filer framework wih incremenal measuremen learning [12], [13]. Among hem, Kwon e al. [13] proposed a paricle filer on a 2D affine Lie group, Aff(2), in a coordinae-invarian way. For 3D visual racking, Pupilli and Calway [14] have shown he possibiliy of applying a paricle filer o 3D edge racking. While [14] demonsraed he racking of simple 3D objecs, Klein and Murray [15] implemened a paricle filering approach which racks complex full 3D objecs in real-ime by exploiing he GPU. Mörwald e al. [16] also exploied he parallel power of GPU o implemen a fas model-based 3D visual racker. Wih edges from 3D CAD, hey also employed edges from exure which possibly conribues o avoid false edge correspondences as well as o enhance he accuracy of pose esimaes. Teulière e al. [17] recenly addressed a similar problem by mainaining muliple hypoheses from low-level edge correspondences. Wih a few excepions [18], mos of he work has made an assumpion in which rackers sar from a given pose. Several effors [15], [14] used annealed paricle filers o find he rue pose from scrach wihou performing an appropriae iniializaion, bu he search space migh be oo large o converge o he rue pose in reasonable ime, and even i migh no converge o he pose afer enough ime elapses. I is hus more desirable o employ oher informaion for iniializaion. The BLORT [18] employed SIFT keypoins [8] o recognize objecs and used hem for paricle iniializaion. In his paper we uilize a paricle filering echnique on he SE(3) group ha is based on [19]. For robus 3D visual racking, we employ keypoin feaures in iniializaion and edges in he calculaion of measuremen likelihoods. Like [15], our sysem can rack complex objecs by performing a self-occlusion es. By mainaining muliple hypoheses, our algorihm can reliably rack an objec on challenging image sequences ha have complex background and heavy cluer. Our key conribuions are as follows: We employ keypoin feaures as addiional visual informaion. While [15], [14] have used annealing paricle filer o find he iniial pose, we iniialize paricles o

2 highly probable saes based on pose esimaes calculaed from keypoin correspondences. The iniialized paricles end o converge faser han he usual annealed paricle filering. While previous edge-based rackers [15], [17] have employed random walk models as a moion model, we apply a firs-order auoregressive (AR) sae dynamics on he SE(3) group o be more effecive. To be fully auomaic and reliable in pracical seings, our approach moniors he number of effecive paricles and use he value o decide when he racker requires re-iniializaion. This paper is organized as follows. In Secion II, we inroduce a paricle filering framework wih sae and measuremen equaions. The AR sae dynamics is hen represened in Secion II-B. Afer explaining how paricles are iniialized and heir likelihoods are evaluaed in Secion II-C and II- D, respecively, he re-iniializaion scheme is represened in II-E. Experimenal resuls on various image sequences are shown in Secion III. II. PARTICLE FILTER ON THE SE(3) GROUP In 3D visual racking, a sae represens a 6-DOF pose of a racked objec, and racking esimaes ime-varying change of coordinaes. I is well known ha he rajecory is no on general vecor space, raher i is on Lie groups in general, he Special Euclidean group SE(3) and he affine group Aff(2) in 3D and 2D visual racking, respecively. Since he rajecory we wan o esimae is on a Lie group, he paricle filer should be applied on Lie groups. Mone Carlo filering on Lie groups is explicily addressed in [2], [19], [13]. As argued in [19] and [13], filering performance and noise disribuion of local coordinae-based paricle filering approaches are dependen on he choice of he local coordinaes, while paricle filering on Lie groups is coordinaeinvarian. A. Sae and Measuremen Equaions From he coninuous general sae equaions on he SE(3) group, discree sysem equaions is acquired via he firsorder exponenial Euler discreizaion [19]: X = X 1 exp(a(x, ) + dw ), (1) 6 dw = ǫ,i E i, i=1 ǫ = (ǫ,1,...,ǫ,6 ) T N( 6 1,Σ w ) where X SE(3) is he sae a ime, A : SE(3) se(3) is a possibly nonlinear map, dw represens he Wiener process noise on se(3) wih a covariance Σ w R 6 6, E i are he i-h basis elemens of se(3). The corresponding measuremen equaion is hen: y = g(x ) + n, n N( Ny 1,Σ n ) (2) where g : X R Ny is a nonlinear measuremen funcion and n is a Gaussian noise wih a covariance Σ n R Ny Ny. B. AR Sae Dynamics The dynamic model for sae evoluion is an essenial par ha has a significan impac on racker performance. However, many paricle filer-based rackers have been based on a random walk model because of is simpliciy [15], [17]. AR sae dynamics is a good alernaive since i is flexible, ye simple o implemen. In (1), he erm A(X, ) deermines he sae dynamics. A rivial case, A(X, ) =, is a random walk model. [13] modeled his via he firs-order AR process on he Aff(2) as: X = X 1 exp(a 1 + dw ), (3) A 1 = alog(x 1 2 X 1) (4) where a is he AR process parameer. Since he SE(3) is a compac conneced Lie group, he AR process model also holds on he SE(3) group [21]. C. Paricle Iniializaion using keypoin Correspondences Mos of he paricle filer-based rackers assume ha iniial saes are given. In pracice, iniial paricles are crucial o ensure convergence o a rue sae. Several rackers [15], [14] search for he rue sae from scrach, bu i is desirable o iniialize paricle saes by using oher informaion. Using keypoins allows for direc esimaion of 3D pose, bu due o he need for a significan number of correspondences i is eiher slow or inaccurae. As such, keypoin correspondences are well suied for he filer iniializaion. For iniializaion, we use so-called keyframes which are composed of 2D images and keypoins coordinaes (2D and 3D) ha have been saved offline. An inpu image coming from a monocular camera is mached wih he keyframes by exracing keypoins and comparing hem. To find keypoin correspondences efficienly, we employ he Bes-Bin-Firs (BBF) algorihm using kd-ree daa srucure [22] ha allows execuion of he search in O(n log n). As described in [8], he raio es is hen performed o find disincive feaure maches. While we used RANSAC [23] afer deermining puaive correspondences in our previous work [24], we skip his procedure because in he paricle filer framework we can iniialize paricles in an alernaive way in which he basic idea is similar o RANSAC. Insead of explicily performing RANSAC, we randomly selec a se of correspondences from he given puaive correspondences and esimae a possible se of poses calculaed from hem. Since we have 3D coordinaes of keypoins in keyframes, we ge 2D-3D correspondences from he maching process described above. So we can regard his problem as he Perspecive-n-Poin (PnP) problem, in which he pose of a calibraed monocular camera is esimaed from n 2D-3D poin correspondences, on each se of correspondences. To find a pose from he correspondences, we use he EPnP algorihm [25] ha provide a O(n) ime non-ieraive soluion for he PnP problem. Afer all paricle poses are iniialized from randomly seleced minimum correspondences, weighs of paricles are assigned from he number of remaining correspondences c r and he number of inlier correspondences c i which coincide wih

3 Algorihm 1 Overall algorihm Iniializaion 1) Se :=. 2) Se number of paricles as N 3) For i := 1,...,N, se X (i) via EPnP, and A (i) := ) For i := 1,...,N, normalize weighs π (i) via (5), by (6). 5) For i := 1,...,N, draw from X (i) according o π (i) o produce X (i). Imporance Sampling 1) Se := ) For i := 1,...,N, draw X (i) P(X X (i) 1,Σ w) by a) Generae he Gaussian ǫ N(,Σ w ), and propagae X (i) 1 b) Compue A (i) o X (i) wih A (i) 1 wih (4). 3) For i := 1,...,N, opimize X (i) via (3). wih IRLS via (9) and (1). 4) For i := 1,...,N, evaluae he imporance weighs o X (i) via (8) and (11). 5) For i := 1,...,N, normalize he imporance weighs π (i) by (12). 6) Evaluae N eff by (13) Resampling 1) If N eff < N hres a) Go o Iniializaion 3) and iniialize X (i), π (i), and A (i) for i := 1,...,N. 2) Oherwise a) For i := 1,...,N, resample from X (i) A (i) and o and A (i) wih probabiliy proporional o π (i) produce i.i.d. random samples X (i) b) For i := 1,...,N, se π (i) := π (i) c) Go o Imporance Sampling. := 1 N. he pose calculaed from he randomly seleced se. For i = 1,...,N where N is he number of paricles, he iniial weighs of paricles are assigned as: p(y X (i) ) e ( λc cr c i ) cr where λ c is a parameer. Then he weighs are normalized by: π (i) = N j=1 π (j) Afer weighs are normalized, paricles are randomly drawn wih probabiliy proporional o hese weighs. By doing so, we can generae probable iniial pose hypoheses. We iniialize paricles when he number of correspondences is bigger or equal o 9, and he number of randomly seleced minimum correspondences is 7.. D. Measuremen Likelihood and Opimizaion using IRLS Once each paricle is iniialized and propagaed according o AR process and Gaussian noise, i has o be evaluaed (5) (6) based on is measuremen likelihood. In edge-based racking, a 3D wireframe model is projeced o a 2D image according o a paricle sae X (i). Then a se of poins is sampled along edges in he wireframe model per a fixed disance. The sampled poins are mached o neares edge pixels from he image by 1D perpendicular search [2], [24]. Then he measuremen likelihood can be calculaed from he raio beween he number of mached sample poins p m and he number of visible sample poins p v which pass a selfocclusion es as: p(y X ) e (pv pm) ( λv ) pv where λ v is a parameer. This likelihood has been similarly used in [15]. Anoher choice is aking arihmeic average disances (error) ē beween he mached sample poins and he edge pixels [17]: p(y X ) e ( λeē) where λ e is also a parameer o be uned. We noiced ha boh likelihoods are valid, and we empirically found ha using boh erms shows beer resuls. Therefore, in our approach he measuremen likelihood is evaluaed as: p(y X ) e (pv pm) ( λv pv ) e ( λeē) One of he challenges in paricle filering for 3D visual racking is he large sae space, so a large number of paricles is usually required for reliable racking performance. To reduce he number of paricles, [15] has used an annealed paricle filer, and [26], [17] have selecively employed local opimizaions in a subse of paricles. For more accurae resuls, we opimize paricles as well, in which Ieraive Reweighed Leas Squares (IRLS) is employed [24], [2]. From IRLS, he opimized paricle X (i) is calculaed as follows: X (i) = X (i) exp( (7) (8) 6 µ i E i ) (9) i=1 µ = (J T WJ) 1 J T We (1) where µ R 6 is he moion velociy ha minimizes he error vecor e R Ny, J R Ny 6 is a Jacobian marix of e wih respec o µ obained by compuing parial derivaives a he curren pose, and W R Ny Ny is a weighed diagonal marix. The diagonal elemen in W is w i = 1 c+e i where c is a consan and e i is i-h elemen of e. (For more deail informaion abou J and IRLS, please refer o [24], [2]). Noe ha he measuremen likelihood in (8) is calculaed before he IRLS opimizaion. To assign weighs of paricles, we have o evaluae he likelihood again wih he opimized sae X (i). However, compuing he likelihood of every paricle again is compuaionally expensive. Since every paricle is opimized a he same ime, we can approximae p(y X (i) ) as: p(y X (i) ) p(y X (i) ) (11)

4 Fig. 1. Tracking resuls wih (yellow wireframe) and wihou (red wireframe) our paricle filer for he four argeed objecs. From op o boom, eabox, book, cup and car door. From lef o righ, < 1, = 1, = 2, = 3, = 4 and = 5 where is he frame number. The very lef images are resuls of he pose iniializaion. Noe ha yellow wireframes are well fied o he racking objecs, while red wireframes are frequenly mislocalized. 1 paricles are used for he paricle filer. (i.e. N = 1) Then he weigh is normalized o π (i) π (i) = N j=1 π (j) E. Re-iniializaion based on N eff by: (12) Ideally a racked objec should be visible during an enire racking session. In realiy, however, i is quie common ha he objec goes ou of frame or is occluded by oher objecs. In hese cases, he racker is required o re-iniialize he racking. In general sequenial Mone Carlo mehods, he effecive paricle size N eff has been inroduced as a suiable measure of degeneracy [27]. Since i is hard o evaluae N eff exacly, an alernaive esimae N eff is defined [27]: N eff = 1 N i=1 ( π(i) ) 2 (13) Ofen i has been used as a measure o execue he resampling procedure. Bu, in our racker we resample paricles every frame, and hence we use N eff as a measure o do reiniializaion. When he number of effecive paricles is below a fixed hreshold N hres, he re-iniializaion procedure is performed. The overall algorihm is shown in Algorihm 1. III. EXPERIMENTAL RESULTS In his secion, we validae our proposed paricle filerbased racker via various experimens. Firs, we compare he performance of our approach wih he previous single hypohesis racker [24] which was based on IRLS. For he comparison, we use new challenging image sequences Teabox Book Cup Car door TABLE I RMS ERRORS IN THE GENERAL TRACKING RMS Errors x y z roll pich yaw The error unis of ranslaion and roaion are meer and degree, respecively. The upper rows are he resuls of he previous approach [24]. The lower rows are he resuls of he proposed approach. In boh upper and lower rows, beer resuls are indicaed in bold numbers. as well as image sequences used in [24]. To verify he effeciveness of he AR sae dynamics, we show he resuls of our proposed racker wih and wihou he AR sae dynamics. A. Experimen 1: Image Sequences from Choi and Chrisensen [24] In [24], he sequences of images were capured from a monocular camera in saic objec and moving camera seing. A se of image sequences used in Secion III-A.1 are acquired o es racking performance in general seing, i.e. no occlusion, relaively simple background, reasonable cluer, and smooh movemens of he camera. To es reiniializaion capabiliy, a sequence of images is capured

5 X Translaion 2 Y Translaion Z Translaion Roll Angle Pich Angle Yaw Angle Ground Truh Fig. 2. Pose plos of he eabox objec in he general racking es. The proposed approach (PF+IRLS) shows superior accuracy han he previous approach (IRLS). Especially, using our paricle filer significanly enhances roaional accuracy. IRLS PF+IRLS 3 Normalized Translaional Residual: T 15 Normalized Roaional Residual: R IRLS PF+IRLS Fig. 3. Normalized residual plos of he eabox objec in he general racking es. In general, he proposed approach (PF+IRLS) shows lower residual and more consisen resuls han he previous one (IRLS). wih fas camera moion and occlusions. This sequence is esed in Secion III-A.2. 1) General Tracking: The single hypohesis racker in [24] has shown reasonable racking resuls on he general racking sequences. To show quaniaive resuls, AR markers are employed o gaher ground ruh pose esimaes. To compare our proposed approach wih he previous one, we execued our new approach on he same image sequences. The racking resuls are shown in Fig. 1. In he proposed approach, 1 paricles are mainained, and he mean of paricles is depiced in he figures. To calculae he mean of paricles, we follow he mean roaion evaluaion of he special orhogonal group SO(3) by Moakher [28] ha is also considered in [19], [17]. To clearly show he difference of he wo approaches, boh of he pose resuls are depiced in he same image sequences. The yellow and red wireframes are projeced o images wih respec o he pose resuls esimaed by he proposed approach and previous approach, respecively. We can easily verify ha he proposed approach shows more accurae resuls. To decompose pose resuls, 6-DOF pose and residual plos of he eabox objec are represened in Fig. 2 and 3, respecively. Based on he plos, we can easily see he difference where he proposed approach (PF+IRLS) shows much beer resuls han he previous approach (IRLS). Noe ha he considerable differences in orienaion esimaes. This is due o some false edge correspondences ha lead o errors in pose, mainly in orienaion. Since our paricle filer considers muliple hypoheses and resamples based on measuremen likelihood, i is quie robus o false edge correspondences from which he single hypohesis racker is ofen suffered. A quaniaive analysis of hese ess is represened in Table I which shows he roo mean square (RMS) errors. For each objec, he upper rows are he resuls of he previous approach and he lower rows are he resuls of he proposed approach. Wih a few excepions, he proposed approach ouperforms he previous one in erms of accuracy. As menioned earlier, he proposed approach shows beer orienaion esimaes. Alhough here are a few excepions, he difference is abou 1 mm in ranslaion and.16 deg in roaion. Since he ground ruh was measured via he AR marker and he displacemen beween he objec and he AR marker was measured manually, ha errors migh come from hese measures. 2) Re-iniializaion: In our previous sysem, we used a simple heurisic in which he difference in posiion of he objec beween frames and he number of valid sample poins are moniored o rigger re-iniializaion [24]. While ha heurisic works well when he pose hypohesis drifs fas, i migh no always be he case when he hypohesis suck in local minima. Here we propose anoher way for reiniializaion by aking advanage of muliple hypoheses. As in Algorihm 1, our sysem re-iniializes when he number of effecive paricles N eff is below a hreshold. To verify his mehod, we run he proposed racker on he re-iniializaion sequence. The racking resuls are shown in Fig. 4 and he number of effecive paricles is ploed over he frame numbers in Fig. 5. The gray line represens he hreshold value N hres. When he racked objec goes ou of frames, images are blurred because of camera shaking, or he objec is occluded wih a paper, he N eff decreases significanly, and ha riggers he re-iniializaion. During re-iniializaion, he racker maches keypoins unil i has a leas he minimum

6 Fig. 4. Tracking resuls on he re-iniializaion sequence. From op-lef o boom-righ, he frame numbers are =, 1, 2, 275, 3, 328, 4, 5, 589, 6, 627, 7, 8, 9, 984, 1, 136, and 11. The green wireframes represen paricles and he yellow hick wireframe shows he mean of paricles on each image. When he racked objec goes ou of frames, images are blurred because of camera shaking, and he objec is occluded wih a paper, our racker re-iniialize. (N = 1) N eff Fig. 5. The N eff plo for he re-iniializaion sequence. (N = 1) number of keypoin correspondences. Once enough poin correspondences are acquired, he proposed sysem iniializes paricles. B. Experimen 2: More Challenging Image Sequences Since he aforemenioned image sequences were prepared for he previous approach, i is relaively easy o rack objecs in hese sequences. Therefore, we need oher daases o compare our new approach. So we prepared more challenging image sequences in which a complex background is considered. 1) Effeciveness of Paricle Filer: When he background is relaively simple, single hypohesis edge-based racking works reasonably. Bu i is quie challenging o reliably rack an objec when he background is complex or here is an amoun of cluer. These challenging siuaions ofen make erraic edge correspondences, hence single hypohesis racking can be a fragile soluion in hese cases. To validae his argumen, we compare our proposed racker wih he single hypohesis racker. We capured wo image sequences for he book and cup objecs. To make he background complex, we pu hese objecs on a camera calibraion plae in which he grid paern is likely o generae false edge correspondences. The comparaive racking resuls are shown in Fig. 6. The grid paern and he background exure play a role as srong cluer, hence he previous approach suffers from he local minima. However, our approach dependably racks objecs in spie of he cluer. 2) Effeciveness of AR Sae Dynamics: To verify he effec of he AR sae dynamics, we execue he proposed approach wih and wihou he AR sae dynamics. To disable he dynamics, we se he parameer a in (4) as which is equivalen o a random walk model. For fair comparison, we use he same parameers excep he AR parameer. We es on a sequence of he book objec used in he previous experimen. The racking resuls are represened in Fig. 7. Alhough boh use he same number of paricles, Gaussian noise, and measuremen likelihood, he racking performances are quie disincive. This difference is mainly due o he AR sae dynamics which propagaes paricles according o he camera moion. IV. CONCLUSIONS We have presened an approach o 3D visual objec racking based on a paricle filering algorihm on he SE(3) group. For fas paricle convergence, we employed keypoin feaures and iniialized paricles by using a linear ime non-ieraive soluion for he PnP problem. Paricles are propagaed by he sae dynamics which is given by an AR process on he SE(3), and he sae dynamics disribued paricles more effecively. Measuremen likelihood was calculaed from boh he residual and he number of valid sample poins of he edge correspondences. During he racking, he proposed sysem appropriaely re-iniialized by iself when he number of effecive paricles is below a hreshold. Our approach has been esed via various experimens in which our muliple hypoheses racker has shown noable performance on challenging background and cluer. One of he possibiliies for fuure work is exploiing he parallel power of GPU for real-ime performance [15], [16]. Anoher ineres is in muliple objec recogniion and racking [29] which is necessary for a realisic scenario.

7 Fig. 6. Tracking resuls on more challenging image sequences. Top and boom rows represen seleced frames from book and cup sequences, respecively. For comparison, he poses esimaed by he proposed (PF+IRLS) and he previous (IRLS) approaches are depiced in yellow and red wireframes, respecively. Because of background exure, he previous approach is ofen misled o local minima, while he proposed approach robusly esimaes poses of objecs under ambiguous background exure and fas camera moions. (N = 1) Fig. 7. Tracking resuls for he proposed approach wih (yellow wireframe) and wihou (red wireframe) he AR sae dynamics. I shows ha he AR sae dynamics conribues o propagae paricles more effecively. (N = 1) V. ACKNOWLEDGMENTS This work was parially funded and developed under a Collaboraive Research Projec beween Georgia Tech and he General Moors R&D, Manufacuring Sysems Research Laboraory. The suppor from General Moors is graefully acknowledged. REFERENCES [1] C. Harris, Tracking wih Rigid Objecs. MIT Press, [2] T. Drummond and R. Cipolla, Real-ime visual racking of complex srucures, PAMI, vol. 24, no. 7, pp , 22. [3] A. I. Compor, E. Marchand, and F. Chaumee, Robus model-based racking for robo vision, in IROS, vol. 1, 24. [4] E. Rosen and T. Drummond, Fusing poins and lines for high performance racking, in ICCV, vol. 2, 25. [5] L. Vacchei, V. Lepei, and P. Fua, Combining edge and exure informaion for real-ime accurae 3d camera racking, in ISMAR, 24, pp [6] M. Pressigou and E. Marchand, Real-ime 3d model-based racking: Combining edge and exure informaion, in ICRA, 26, pp [7] G. Klein and T. Drummond, Tighly inegraed sensor fusion for robus visual racking, Image and Vision Compuing, vol. 22, no. 1, pp , 24. [8] D. G. Lowe, Disincive image feaures from scale-invarian keypoins, IJCV, vol. 6, no. 2, pp , 24. [9] H. Bay, A. Ess, T. Tuyelaars, and L. V. Gool, Speeded-up robus feaures (SURF), Compuer Vision and Image Undersanding, vol. 11, no. 3, pp , 28. [1] C. Kemp and T. Drummond, Dynamic measuremen clusering o aid real ime racking, in ICCV, 25, pp [11] M. Isard and A. Blake, Condensaion condiional densiy propagaion for visual racking, IJCV, vol. 29, no. 1, pp. 5 28, [12] D. A. Ross, J. Lim, R. S. Lin, and M. H. Yang, Incremenal learning for robus visual racking, IJCV, vol. 77, no. 1, pp , 28. [13] J. Kwon and F. C. Park, Visual racking via paricle filering on he affine group, IJRR, vol. 29, no. 2-3, p. 198, 21. [14] M. Pupilli and A. Calway, Real-ime camera racking using known 3D models and a paricle filer, in ICPR, vol. 1, 26. [15] G. Klein and D. Murray, Full-3d edge racking wih a paricle filer, BMVC, 26. [16] T. Mörwald, M. Zillich, and M. Vincze, Edge racking of exured objecs wih a recursive paricle filer, in 19h Inernaional Conference on Compuer Graphics and Vision (Graphicon), Moscow, 29, pp [17] C. Teulière, E. Marchand, and L. Eck, Using muliple hypohesis in model-based racking, in ICRA, 21. [18] T. Mörwald, J. Prankl, A. Richsfeld, M. Zillich, and M. Vincze, BLORT - he blocks world roboic vision oolbox, in In Bes Pracice in 3D Percepion and Modeling for Mobile Manipulaion (in conjuncion wih ICRA 21), 21. [19] J. Kwon, M. Choi, F. C. Park, and C. Chun, Paricle filering on he Euclidean group: framework and applicaions, Roboica, vol. 25, no. 6, pp , 27. [2] A. Chiuso and S. Soao, Mone Carlo filering on Lie groups, in IEEE Conference on Decision and Conrol, vol. 1, 2, pp [21] J. Xavier and J. H. Manon, On he generalizaion of AR processes o Riemannian manifolds, Proceedings of IEEE Inernaional Conference on Acousics, Speech, and Signal Processing., 26. [22] J. Beis and D. Lowe, Shape indexing using approximae nearesneighbour search in high-dimensional spaces, in CVPR, 1997, pp [23] M. A. Fischler and R. C. Bolles, Random sample consensus: a paradigm for model fiing wih applicaions o image analysis and auomaed carography, Commun. ACM, vol. 24, no. 6, pp , [24] C. Choi and H. I. Chrisensen, Real-ime 3D model-based racking using edge and keypoin feaures for roboic manipulaion, in ICRA, 21, pp [25] V. Lepei, F. Moreno-Noguer, and P. Fua, EPnP: An accurae O(n) soluion o he PnP problem, IJCV, vol. 81, no. 2, pp , 29. [26] M. Bray, E. Koller-Meier, and L. V. Gool, Smar paricle filering for 3D hand racking, in Sixh IEEE Inernaional Conference on Auomaic Face and Gesure Recogniion, 24. Proceedings, 24, pp [27] A. Douce, S. Godsill, and C. Andrieu, On sequenial Mone Carlo sampling mehods for Bayesian filering, Saisics and compuing, vol. 1, no. 3, pp , 2. [28] M. Moakher, Means and averaging in he group of roaions, SIAM Journal on Marix Analysis and Applicaions, vol. 24, no. 1, p. 116, 23. [29] A. Colle, D. Berenson, S. S. Srinivasa, and D. Ferguson, Objec recogniion and full pose regisraion from a single image for roboic manipulaion, in ICRA, 29, 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

RGB-D Object Tracking: A Particle Filter Approach on GPU

RGB-D Object Tracking: A Particle Filter Approach on GPU RGB-D Objec Tracking: A Paricle Filer Approach on GPU Changhyun Choi and Henrik I. Chrisensen Cener for Roboics & Inelligen Machines College of Compuing Georgia Insiue of Technology Alana, GA 3332, USA

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

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

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

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

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

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

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

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

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

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

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

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

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

4.1 3D GEOMETRIC TRANSFORMATIONS

4.1 3D GEOMETRIC TRANSFORMATIONS MODULE IV MCA - 3 COMPUTER GRAPHICS ADMN 29- Dep. of Compuer Science And Applicaions, SJCET, Palai 94 4. 3D GEOMETRIC TRANSFORMATIONS Mehods for geomeric ransformaions and objec modeling in hree dimensions

More information

CENG 477 Introduction to Computer Graphics. Modeling Transformations

CENG 477 Introduction to Computer Graphics. Modeling Transformations CENG 477 Inroducion o Compuer Graphics Modeling Transformaions Modeling Transformaions Model coordinaes o World coordinaes: Model coordinaes: All shapes wih heir local coordinaes and sies. world World

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

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

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

Sam knows that his MP3 player has 40% of its battery life left and that the battery charges by an additional 12 percentage points every 15 minutes.

Sam knows that his MP3 player has 40% of its battery life left and that the battery charges by an additional 12 percentage points every 15 minutes. 8.F Baery Charging Task Sam wans o ake his MP3 player and his video game player on a car rip. An hour before hey plan o leave, he realized ha he forgo o charge he baeries las nigh. A ha poin, he plugged

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

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

NEWTON S SECOND LAW OF MOTION

NEWTON S SECOND LAW OF MOTION Course and Secion Dae Names NEWTON S SECOND LAW OF MOTION The acceleraion of an objec is defined as he rae of change of elociy. If he elociy changes by an amoun in a ime, hen he aerage acceleraion during

More information

Gauss-Jordan Algorithm

Gauss-Jordan Algorithm Gauss-Jordan Algorihm The Gauss-Jordan algorihm is a sep by sep procedure for solving a sysem of linear equaions which may conain any number of variables and any number of equaions. The algorihm is carried

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

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

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

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

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

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

In fmri a Dual Echo Time EPI Pulse Sequence Can Induce Sources of Error in Dynamic Magnetic Field Maps

In fmri a Dual Echo Time EPI Pulse Sequence Can Induce Sources of Error in Dynamic Magnetic Field Maps In fmri a Dual Echo Time EPI Pulse Sequence Can Induce Sources of Error in Dynamic Magneic Field Maps A. D. Hahn 1, A. S. Nencka 1 and D. B. Rowe 2,1 1 Medical College of Wisconsin, Milwaukee, WI, Unied

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

MOTION DETECTORS GRAPH MATCHING LAB PRE-LAB QUESTIONS

MOTION DETECTORS GRAPH MATCHING LAB PRE-LAB QUESTIONS NME: TE: LOK: MOTION ETETORS GRPH MTHING L PRE-L QUESTIONS 1. Read he insrucions, and answer he following quesions. Make sure you resae he quesion so I don hae o read he quesion o undersand he answer..

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

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

Low-Cost WLAN based. Dr. Christian Hoene. Computer Science Department, University of Tübingen, Germany

Low-Cost WLAN based. Dr. Christian Hoene. Computer Science Department, University of Tübingen, Germany Low-Cos WLAN based Time-of-fligh fligh Trilaeraion Precision Indoor Personnel Locaion and Tracking for Emergency Responders Third Annual Technology Workshop, Augus 5, 2008 Worceser Polyechnic Insiue, Worceser,

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

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

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

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

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

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

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

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

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

Real-Time Avatar Animation Steered by Live Body Motion

Real-Time Avatar Animation Steered by Live Body Motion Real-Time Avaar Animaion Seered by Live Body Moion Oliver Schreer, Ralf Tanger, Peer Eiser, Peer Kauff, Bernhard Kaspar, and Roman Engler 3 Fraunhofer Insiue for Telecommunicaions/Heinrich-Herz-Insiu,

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

Motor Control. 5. Control. Motor Control. Motor Control

Motor Control. 5. Control. Motor Control. Motor Control 5. Conrol In his chaper we will do: Feedback Conrol On/Off Conroller PID Conroller Moor Conrol Why use conrol a all? Correc or wrong? Supplying a cerain volage / pulsewidh will make he moor spin a a cerain

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

RGBD Data Based Pose Estimation: Why Sensor Fusion?

RGBD Data Based Pose Estimation: Why Sensor Fusion? 18h Inernaional Conference on Informaion Fusion Washingon, DC - July 6-9, 2015 RGBD Daa Based Pose Esimaion: Why Sensor Fusion? O. Serdar Gedik Deparmen of Compuer Engineering, Yildirim Beyazi Universiy,

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

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

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 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

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

Stereo Vision Based Navigation of a Six-Legged Walking Robot in Unknown Rough Terrain

Stereo Vision Based Navigation of a Six-Legged Walking Robot in Unknown Rough Terrain Sereo Vision Based Navigaion of a Six-Legged Walking Robo in Unknown Rough Terrain Anne Selzer, Heiko Hirschmüller, Marin Görner Absrac This paper presens a visual navigaion algorihm for he six-legged

More information

Motion Estimation of a Moving Range Sensor by Image Sequences and Distorted Range Data

Motion Estimation of a Moving Range Sensor by Image Sequences and Distorted Range Data Moion Esimaion of a Moving Range Sensor by Image Sequences and Disored Range Daa Asuhiko Banno, Kazuhide Hasegawa and Kasushi Ikeuchi Insiue of Indusrial Science Universiy of Tokyo 4-6-1 Komaba, Meguro-ku,

More information

It is easier to visualize plotting the curves of cos x and e x separately: > plot({cos(x),exp(x)},x = -5*Pi..Pi,y = );

It is easier to visualize plotting the curves of cos x and e x separately: > plot({cos(x),exp(x)},x = -5*Pi..Pi,y = ); Mah 467 Homework Se : some soluions > wih(deools): wih(plos): Warning, he name changecoords has been redefined Problem :..7 Find he fixed poins, deermine heir sabiliy, for x( ) = cos x e x > plo(cos(x)

More information

arxiv: v1 [cs.cv] 11 Jan 2019

arxiv: v1 [cs.cv] 11 Jan 2019 A General Opimizaion-based Framewor for Local Odomery Esimaion wih Muliple Sensors Tong Qin, Jie Pan, Shaozu Cao, and Shaojie Shen arxiv:191.3638v1 [cs.cv] 11 Jan 219 Absrac Nowadays, more and more sensors

More information

STRING DESCRIPTIONS OF DATA FOR DISPLAY*

STRING DESCRIPTIONS OF DATA FOR DISPLAY* SLAC-PUB-383 January 1968 STRING DESCRIPTIONS OF DATA FOR DISPLAY* J. E. George and W. F. Miller Compuer Science Deparmen and Sanford Linear Acceleraor Cener Sanford Universiy Sanford, California Absrac

More information

MATH Differential Equations September 15, 2008 Project 1, Fall 2008 Due: September 24, 2008

MATH Differential Equations September 15, 2008 Project 1, Fall 2008 Due: September 24, 2008 MATH 5 - Differenial Equaions Sepember 15, 8 Projec 1, Fall 8 Due: Sepember 4, 8 Lab 1.3 - Logisics Populaion Models wih Harvesing For his projec we consider lab 1.3 of Differenial Equaions pages 146 o

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

Large-scale 3D Outdoor Mapping and On-line Localization using 3D-2D Matching

Large-scale 3D Outdoor Mapping and On-line Localization using 3D-2D Matching Large-scale 3D Oudoor Mapping and On-line Localizaion using 3D-D Maching Takahiro Sakai, Kenji Koide, Jun Miura, and Shuji Oishi Absrac Map-based oudoor navigaion is an acive research area in mobile robos

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

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

A High-Speed Adaptive Multi-Module Structured Light Scanner

A High-Speed Adaptive Multi-Module Structured Light Scanner A High-Speed Adapive Muli-Module Srucured Ligh Scanner Andreas Griesser 1 Luc Van Gool 1,2 1 Swiss Fed.Ins.of Techn.(ETH) 2 Kaholieke Univ. Leuven D-ITET/Compuer Vision Lab ESAT/VISICS Zürich, Swizerland

More information

FIELD PROGRAMMABLE GATE ARRAY (FPGA) AS A NEW APPROACH TO IMPLEMENT THE CHAOTIC GENERATORS

FIELD PROGRAMMABLE GATE ARRAY (FPGA) AS A NEW APPROACH TO IMPLEMENT THE CHAOTIC GENERATORS FIELD PROGRAMMABLE GATE ARRAY (FPGA) AS A NEW APPROACH TO IMPLEMENT THE CHAOTIC GENERATORS Mohammed A. Aseeri and M. I. Sobhy Deparmen of Elecronics, The Universiy of Ken a Canerbury Canerbury, Ken, CT2

More information

Multiple View Discriminative Appearance Modeling with IMCMC for Distributed Tracking

Multiple View Discriminative Appearance Modeling with IMCMC for Distributed Tracking 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 {sanhosh,manj}@ece.ucsb.edu

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

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

4. Minimax and planning problems

4. Minimax and planning problems CS/ECE/ISyE 524 Inroducion o Opimizaion Spring 2017 18 4. Minima and planning problems ˆ Opimizing piecewise linear funcions ˆ Minima problems ˆ Eample: Chebyshev cener ˆ Muli-period planning problems

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

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

Improving Accuracy of Inertial Measurement Units using Support Vector Regression

Improving Accuracy of Inertial Measurement Units using Support Vector Regression Improving Accuracy of Inerial Measuremen Unis using Suppor Vecor Regression Saran Ahuja, Wisi Jiraigalachoe, and Ar Tosborvorn Absrac Inerial measuremen uni (IMU) is a sensor ha measures acceleraion and

More information

Tracking Deforming Objects Using Particle Filtering for Geometric Active Contours

Tracking Deforming Objects Using Particle Filtering for Geometric Active Contours 1470 IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 29, NO. 8, AUGUST 2007 Tracking Deforming Objecs Using Paricle Filering for Geomeric Acive Conours Yogesh Rahi, Member, IEEE, NamraaVaswani,

More information

M(t)/M/1 Queueing System with Sinusoidal Arrival Rate

M(t)/M/1 Queueing System with Sinusoidal Arrival Rate 20 TUTA/IOE/PCU Journal of he Insiue of Engineering, 205, (): 20-27 TUTA/IOE/PCU Prined in Nepal M()/M/ Queueing Sysem wih Sinusoidal Arrival Rae A.P. Pan, R.P. Ghimire 2 Deparmen of Mahemaics, Tri-Chandra

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

Multi-Vehicle Cooperative Local Mapping: A Methodology Based on Occupancy Grid Map Merging

Multi-Vehicle Cooperative Local Mapping: A Methodology Based on Occupancy Grid Map Merging Muli-Vehicle Cooperaive Local Mapping: Mehodology ased on Occupancy Grid Map Merging Hao Li, Manabu Tsukada, Fawzi Nashashibi, Michel Paren To cie his version: Hao Li, Manabu Tsukada, Fawzi Nashashibi,

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

LAMP: 3D Layered, Adaptive-resolution and Multiperspective Panorama - a New Scene Representation

LAMP: 3D Layered, Adaptive-resolution and Multiperspective Panorama - a New Scene Representation Submission o Special Issue of CVIU on Model-based and Image-based 3D Scene Represenaion for Ineracive Visualizaion LAMP: 3D Layered, Adapive-resoluion and Muliperspecive Panorama - a New Scene Represenaion

More information

arxiv: v1 [cs.cv] 11 Jan 2019

arxiv: v1 [cs.cv] 11 Jan 2019 A General Opimizaion-based Framework for Global Pose Esimaion wih Muliple Sensors arxiv:191.3642v1 [cs.cv] 11 Jan 219 Tong Qin, Shaozu Cao, Jie Pan, and Shaojie Shen Absrac Accurae sae esimaion is a fundamenal

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

User Adjustable Process Scheduling Mechanism for a Multiprocessor Embedded System

User Adjustable Process Scheduling Mechanism for a Multiprocessor Embedded System Proceedings of he 6h WSEAS Inernaional Conference on Applied Compuer Science, Tenerife, Canary Islands, Spain, December 16-18, 2006 346 User Adjusable Process Scheduling Mechanism for a Muliprocessor Embedded

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

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

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

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

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

DAGM 2011 Tutorial on Convex Optimization for Computer Vision

DAGM 2011 Tutorial on Convex Optimization for Computer Vision DAGM 2011 Tuorial on Convex Opimizaion for Compuer Vision Par 3: Convex Soluions for Sereo and Opical Flow Daniel Cremers Compuer Vision Group Technical Universiy of Munich Graz Universiy of Technology

More information

Dynamic Route Planning and Obstacle Avoidance Model for Unmanned Aerial Vehicles

Dynamic Route Planning and Obstacle Avoidance Model for Unmanned Aerial Vehicles Volume 116 No. 24 2017, 315-329 ISSN: 1311-8080 (prined version); ISSN: 1314-3395 (on-line version) url: hp://www.ijpam.eu ijpam.eu Dynamic Roue Planning and Obsacle Avoidance Model for Unmanned Aerial

More information

A Fast Non-Uniform Knots Placement Method for B-Spline Fitting

A Fast Non-Uniform Knots Placement Method for B-Spline Fitting 2015 IEEE Inernaional Conference on Advanced Inelligen Mecharonics (AIM) July 7-11, 2015. Busan, Korea A Fas Non-Uniform Knos Placemen Mehod for B-Spline Fiing T. Tjahjowidodo, VT. Dung, and ML. Han Absrac

More information

Terrain Based GPS Independent Lane-Level Vehicle Localization using Particle Filter and Dead Reckoning

Terrain Based GPS Independent Lane-Level Vehicle Localization using Particle Filter and Dead Reckoning Terrain Based GPS Independen -Level Vehicle Localizaion using Paricle Filer and Dead Reckoning Hamad Ahmed and Muhammad Tahir Deparmen of Elecrical Engineering, Lahore Universiy of Managemen Sciences,

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

Reasoning About Liquids via Closed-Loop Simulation

Reasoning About Liquids via Closed-Loop Simulation Reasoning Abou Liquids via Closed-Loop Simulaion Connor Schenck and Dieer Fox Absrac Simulaors are powerful ools for reasoning abou a robo s ineracions wih is environmen. However, when simulaions diverge

More information

AML710 CAD LECTURE 11 SPACE CURVES. Space Curves Intrinsic properties Synthetic curves

AML710 CAD LECTURE 11 SPACE CURVES. Space Curves Intrinsic properties Synthetic curves AML7 CAD LECTURE Space Curves Inrinsic properies Synheic curves A curve which may pass hrough any region of hreedimensional space, as conrased o a plane curve which mus lie on a single plane. Space curves

More information

Computer representations of piecewise

Computer representations of piecewise Edior: Gabriel Taubin Inroducion o Geomeric Processing hrough Opimizaion Gabriel Taubin Brown Universiy Compuer represenaions o piecewise smooh suraces have become vial echnologies in areas ranging rom

More information

1.4 Application Separable Equations and the Logistic Equation

1.4 Application Separable Equations and the Logistic Equation 1.4 Applicaion Separable Equaions and he Logisic Equaion If a separable differenial equaion is wrien in he form f ( y) dy= g( x) dx, hen is general soluion can be wrien in he form f ( y ) dy = g ( x )

More information

Image-Based Pose Estimation for 3-D Modeling in Rapid, Hand-Held Motion

Image-Based Pose Estimation for 3-D Modeling in Rapid, Hand-Held Motion 2011 IEEE Inernaional Conference on Roboics and Auomaion Shanghai Inernaional Conference Cener May 9-13, 2011, Shanghai, China Image-Based Pose Esimaion for 3-D Modeling in Rapid, Hand-Held Moion Klaus

More information

SLAM in Large Indoor Environments with Low-Cost, Noisy, and Sparse Sonars

SLAM in Large Indoor Environments with Low-Cost, Noisy, and Sparse Sonars SLAM in Large Indoor Environmens wih Low-Cos, Noisy, and Sparse Sonars Teddy N. Yap, Jr. and Chrisian R. Shelon Deparmen of Compuer Science and Engineering Universiy of California, Riverside, CA 92521,

More information