A Probabilistic Method for Aligning and Merging Range Images with Anisotropic Error Distribution

Size: px
Start display at page:

Download "A Probabilistic Method for Aligning and Merging Range Images with Anisotropic Error Distribution"

Transcription

1 A Probablstc Method for Algnng and Mergng Range Images wth Ansotropc Error Dstrbuton Ryusuke Sagawa Nanaho Osawa Yasush Yag The Insttute of Scentfc and Industral Research, Osaka Unversty 8-1 Mhogaoka, Ibarak, Osaka, , JAPAN Abstract Ths paper descrbes a probablstc method of algnng and mergng range mages. We formulate these ssues as problems of estmatng the mamum lkelhood. By eamnng the error dstrbuton of a range fnder, we model t as a normal dstrbuton along the lne of sght. To algn range mages, our method estmates the parameters based on the Epectaton Mamzaton (EM) approach. By assumng the error model, the algorthm s mplemented as an etenson of the Iteratve Closest Pont (ICP) method. For mergng range mages, our method computes the sgned dstances by fndng the dstances of mamum lkelhood. Snce our proposed method uses multple correspondences for each verte of the range mages, errors after algnng and mergng range mages are less than those of earler methods that use one-to-one correspondences. Fnally, we tested and valdated the effcency of our method by smulaton and on real range mages. 1. Introducton Many researchers have studed modelng shapes of real world objects by scannng them usng three dmensonal dgtzers such as laser range fnders [5, 11, 3] and structured-lght range fnders [22]. For eample, a major subject of research s the modelng of cultural hertage objects [12, 9] because they are seen to be canddates havng the worthest shapes and appearances for modelng. 3D modelng of the shape of an object s accomplshed by followng three steps: 1. Acqurng the range mages (scannng). 2. Algnng the acqured range mages from dfferent vewponts (algnng). 3. Reconstructng a unfed 3D mesh model (mergng). In the frst step, a target object s observed from varous vewponts. In the second step, multple range mages are algned nto a common coordnate system usng regstraton algorthms that establsh pont correspondences and mnmze the total dstance between those ponts; usng, for eample, feature-based [24, 10], or Iteratve Closest Pont (ICP)-based [1, 29, 2, 14, 17, 18] methods. The thrd step s to merge multple pre-algned range mages. Several approaches have been proposed; for eample, mesh-based [25, 23] and volume-based [8, 4, 26, 19] methods. When algnng and mergng multple range mages t s necessary to fnd correspondng ponts between them. There s some varaton among the methods used. WthICPbased methods, the correspondng ponts are defned by the closest Eucldean dstance [1, 29], and the projecton of the source pont onto the destnaton mesh [2, 14, 17, 15]. If an Eucldean dstance s used, t s equvalent to assumng that the error dstrbuton of a range mage s sotropc. However, the error dstrbuton of an actual range fnder does not seem sotropc. If the dstance s computed by projectng the source pont onto the destnaton mesh, the error s dstrbuted on the lne of projecton. It seems plausble because the error dstrbuton of a range fnder s along the lne of sght, especally for a laser range fnder. However, as these methods fnd one-to-one correspondences, t s dffcult to fnd a correct correspondence when the measurement error of a range fnder s large, even f the ntal poston of range mages s a suffcently good guess. The methods for mergng range mages also fnd correspondng ponts by usng the closest Eucldean dstance [25, 8, 26, 19], and projectng a pont onto a mesh [4]. For the same reason as n the case of algnng range mages, t s dffcult to fnd a correct correspondence when the measurement error of a range fnder s large, and the assumpton of the error dstrbuton s far from the actual dstrbuton. Some approaches that model the error dstrbuton of the range measurement have been proposed. Okatan and Deguch[16] proposed a method to algn range mages wth an error model of range mages and modeled t as a normal dstrbuton along the lne of sght. However, they determnstcally computed the one-to-one correspondng ponts. Wllams and Bennamoun [28] proposed a proba- 1

2 blstc method to algn range mages. Though they modeled the error dstrbuton by a covarance matr, they also used determnstc correspondences. Sagawa et al. [20] proposed a method to remove the nose n range mages along the lne of sght, but they dd not smultaneously consder the error dstrbuton of multple range mages. If we assume that the error dstrbuton of a range fnder s along the lne of sght, the nose of each pel can be reduced by smple temporal averagng by obtanng multple range mages from the same vewpont. However, t s not applcable n the case that the sensor or object moves when acqurng those range mages. Our proposed method can be appled even f the sensor s movng, and only one range mage s obtaned for each vew drecton. In ths paper, we propose probablstc methods to algn and merge range mages. We regard algnng range mages as problems of fndng optmal parameters of mamum lkelhood. Our mergng method fnds a surface of mamum lkelhood durng the converson from range mages to a volumetrc representaton. Because we analyze the error dstrbuton of a laser range fnder, we smplfy the model for the measurement error and usng an ansotropc error model we propose algorthms for algnng and mergng range mages. In Secton 2, we formulate the problems as mamum lkelhood estmates. Then, we analyze the error dstrbuton and propose practcal algorthms n Secton 3. We evaluate our method n Secton 4 and fnally summarze ths paper n Secton Probablstc Method for Algnng and Mergng Range Images Ths secton descrbes a method to algn and merge range mages based on a probablstc framework. Frst, we defne a probablstc model of the error dstrbuton of a range mage that s obtaned by a range fnder. Net, we defne the problems wth the model of error dstrbuton as follows: Algnng range mages has the problem that estmates the parameters that have mamum lkelhood. Mergng range mages has the problem that fnds a surface that has a mamum lkelhood Probablstc Model of Error Dstrbuton n Range Measurement A range mage s a set of 3D ponts acqured by a range fnder and whch are connected f ther postons are close to each other [25]. Therefore, a range mage s represented as a mesh model, consstng of vertces and patches. Because of the nose n range measurement, a 3D pont n a range mage has an error. Fgure 1 shows an eample where there are a true surface S and a observed range mage A. s a pont of A. If s a measurement of a pont s of S, the probablty s represented as a posteror probablty P ( s). If a general model of error dstrbuton s assumed, can surface S range mage A s P( s) Fgure 1. A range mage A s observed by scannng atruesurfaces. A pont of A s a measurement of a pont s of S. The posteror probablty s P ( s). range mage B surface S range mage A s P(s B) P( S) Fgure 2. Two range mages A and B are two samples of observng a true surface S. If B s observed, the probablty of S becomes P (S B) = P (s B), s and the probablty that A s observed s P (A S) = P ( S) f S ests. becausedbyallpartsofs. Thus, we denote the posteror probablty as P ( S) and the probablty of a whole range mage A becomes P (A S) = P ( S). (1) 2.2. Algnng Range Images by Mamum Lkelhood Estmaton Fgure 2 shows a stuaton where there are two range mages A, B and a true surface S. Now, we consder the algnment of the range mage A to B by changng the parameters of rotaton R and by translaton t of the range mage A. In a probablstc framework, the parameter s computed by mamzng the probablty P (A B; θ), whereθ s a vector that represents the parameters R and t. P (A B; θ) s computed as follows: P (A B; θ) = P (A S; θ)p (S B)dS, (2) S Ω where Ω represents the scope of the shape of the true surface S. Usng the Epectaton Mamzaton (EM) approach, the parameter θ that mamzes P (A B; θ) s computed by estmatng the followng condtonal epectaton E[log P (S B) A] = P (S A; θ)logp (S B)dS. (3) S Ω 2

3 If S conssts of n ponts s 1, s 2,, s n, (3) becomes P (s j A; θ)log P (s B)ds 1 ds n s 1 s n j = P (s j A; θ) L(s B)ds 1 ds n (4) s 1 s n j where L(s B) s the log lkelhood log P (s S). Because s P (s A; θ) =1for all =1,,n, t s smplfed as E[log P (S B) A] = P (s A; θ)l(s B)ds. (5) s Snce an actual range fnder has a characterstc error dstrbuton, we propose an algorthm to compute and mamze (5) based on the error dstrbuton n Secton Mergng Range Images by Fndng Surface of Mamum Lkelhood To merge range mages, we compute a sgned dstance feld (SDF) as an ntermedate representaton. Snce t s an mplct representaton, the merged surface becomes a set of ponts that satsfes f() = 0,wheref() s a sgned dstance at. The mplct surface represented by a SDF s converted to a mesh model by the Marchng Cubes algorthm [13]. In ths secton, we propose a method to merge range mages by computng a sgned dstance f() of mamum lkelhood from multple range mages. Fgure 3 shows an eample where there are two range mages A, B, a true surface S and a pont. D () s the dstance from to a pont s of S. The probablty functon for D () d s computed by P (D () d) = P (s A, B; θ)ds, (6) s d where θ s the parameter of algnment. Thus, P (D() = d) s computed by takng average of all neghbor ponts as follows: P (D() =d) = N P (D() d δ) P (D() d+δ) lm δ 0 2δN, (7) where N s the number of neghbor ponts. Snce the magntude of a sgned dstance f() s computed as the dstance from to the nearest neghbor pont of the surface [19], then f s s the nearest neghbor pont, the other ponts, for eample s k n Fgure 3, are farther than s from. Snce the probablty P 1 (D() =d) that only one pont s wthn [d δ, d + δ) s lm δ 0 P (D() d δ) P (D() d + δ), (8) 2δ the probablty densty functon for f() = d s computed by P ( f() = d) =P (D() =d) P 1 (D() =d) (9) By fndng the mamum lkelhood of P ( f() = d), we determne the magntude of the sgned dstance. range mage B surface S s k D () k range mage A P(s A,B) s D () Fgure 3. An eample of mergng two range mages. D () s the dstance from to a pont s of a true surface S. The remanng ssue s to determne the sgn of f(). In [19], the sgn s determned by consderng the angle between p and the normal vector n at p, f the nearest pont s p. Namely, the sgn of f() becomes postve f (p ) n < 0, and negatve otherwse. In our method, f the magntude of f() s d, we determne the sgn by computng a weghted sum of the probablty: W (,d)= P (s p)sgn((p ) n)dpds, p s =d (10) where p s a pont of the range mages and sgn(a) s 1 f a<0and negatve otherwse. Therefore, the sgn of f() becomes sgn(w (,d)). In ths secton, we eplan our method for mergng just two range mages; however, t can easly be etended to cases of three or more range mages. 3. Algorthm wth Ansotropc Error Model Ths secton descrbes a practcal algorthm to algn and merge range mages. Snce t s dffcult to compute the probabltes descrbed n Secton 2 for general probablstc models, we frst analyze the error dstrbuton of an actual laser range fnder and smplfy the model of error dstrbuton. Net, we propose algorthms based on the model Smplfed Model of Ansotropc Error Dstrbuton Here, we used a Canesta DP200 [3] as the laser range fnder to analyze the error dstrbuton. As a prelmnary eperment, we evaluated the error dstrbuton of the sensor by observng a small object. Snce t has a lght source and a camera, t obtans a 2D range mage by observng the reflected lght. The poston and sze of the small object s adjusted so that t s projected wthn a pel of the mage. Fgure 4 shows the results of the measurements. Fgure 4(a) s the hstogram of the number of measurements for each pel along the horzontal as of the mage. Snce most of the measurements are projected onto only one pel, 3

4 Number of measurements Number of measurements X as (pel) (a) Dstance (mm) (b) Fgure 4. Hstograms of measurng a small object: (a) along the horzontal as of a range mage, (b) along the lne of sght. t can be assumed that a pel s not affected by other pels, whch s reasonable f an object s n focus. It s equally reasonable to assume that the error ests only along the vew drecton. Fgure 4(b) s the hstogram of the number of measurements for a pel along the lne of sght. Snce the dstrbuton s smlar to a normal dstrbuton, we assume that t s a normal dstrbuton N(d, σ 2 ),whered s the dstance to the true surface and σ s the standard devaton. Snce [21] analyzed for the standard devaton of the error, we use the result. Consequently, snce we assume that the error only ests along the vew drecton, the model of error dstrbuton s smplfed to a 1D normal dstrbuton. Though ths error model s valdated only for the Canesta DP200, t can be appled to other range fnders, especally laser range fnders Algnng Range Images Fgure 5 shows a stuaton where there are two range mages A, B to be algned and an assumed true surface S. If we assume the model of error dstrbuton descrbed n Secton 3.1, a verte of the range mage A s a measurement of s of the surface S, whch s on the same lne of sght of the range mage A. The probablty of becomes P ( s )= 1 ep( s 2 2πσ 2σ 2 ). (11) Meanwhle, the measurement of s n the range mage B s y, whch s on the same lne of sght of B. The probablty of y s smlar to (11). Snce and y only depend on s, range mage B surface lne of sght of B y S range mage A P(y s ) s P( s ) lne of sght of A Fgure 5. Correspondence ponts between range mages A, B and an assumed true surface S. s s on the lnes of sght of both and y. (5) becomes E[log P (S B) A] = s P (s ; θ)l(s y )ds. (12) In the mamzaton step of the EM algorthm, the estmated parameter ˆθ becomes by Bayes s rule ˆθ = arg θ ma P (s ; θ)l(s y )ds s = arg θ ma P ( s ; θ)l(y s )ds,(13) s where we drop P (s ),P( ) and P (y ) because they are constant wth respect to the parameters f we assume that there s no constrant on the shape of S. To resolve (13), we compute the weghted sum of several samples of s nstead of computng the ntegral of s as shown n Fgure 6. Snce the error dstrbuton s a normal dstrbuton, t becomes ˆθ =arg θ mn j P (s,j ) y,j s,j 2, (14) where s,j sasampleofs and y,j s the correspondng pont of s,j. The dstrbuton of s,j s gven by user. P (s,j ) s computed by 1 2 erf(t j + t j+1 2 2σ ) erf(t j + t j 1 2 ), (15) 2σ where erf(z) = 2 z π 0 e t2 dt, ands j = + t j v (v s the unt vector along the lne of sght). Therefore, the EM algorthm teratvely mnmzes the dstance of the correspondng ponts. Ths s smlar to the ICP-based methods, especally [14, 15], whch fnd correspondng ponts along the vew drecton. However, snce they do not consder the error dstrbuton, they use only the correspondence between and y 3 n the case of Fgure 6. Thus, our algorthm s the same as [15] for mnmzng an energy functon, but t uses a dfferent energy functon (14). 4

5 y,4 y,3 y,5 y,2 range mage A y,1 range mage B v s,1 s,2 s,3 s,4 s,5 Probablty P(D()=d_k) P_1(D()=d_k) P b k Dstance d_k Probablty P_b_k Fgure 6. Computng the weght sums of samples of s,j nstead of computng the ntegral of s Mergng Range Images The net step s to merge range mages. To fnd the mamum of the probablty of P ( f() = d), we create a hstogram nstead of computng the lmt of the probablty epressed n (9). Our algorthm to compute the sgned dstance f() s as follows: 1. Fnd several neghbor ponts of range mages whose dstances from are smaller than d mn +3σ, where d mn s the dstance to the nearest neghbor pont and σ s the standard devaton of measurement. 2. Compute P (D () d k ) for k =1, 2,...n for each found neghbor pont, where the bn of a hstogram s defned between d k and d k Compute the sum and product of the probablty functon for all neghbor ponts: P (D() =d k )= {P (D() d k+1) P (D() d k )} P 1(D() =d k )= P (D() d k+1) P (D() d k) 4. Fnd the bn b k that has the mamum value of P bk = P (D() =d k ) P 1 (D() =d k ). 5. Interpolate the dstance f() of the mamum probablty by fttng a quadratc functon to P bk 1,P bk and P bk The sgn s determned by computng (10) for the bn b k. To fnd the neghbor ponts, we use a k-d tree [7] to reduce the cost wth a small modfcaton to fnd ponts whose dstances are wthn d mn +3σ. Snce other ponts of the range mages farther than d mn +3σ do not contrbute to the probablty P (D() =d k ), we omt them from the computaton. The wdth of a bn s defned by the user because t s a trade-off between cost and accuracy. Snce our algorthm s based on [19], the poston of s determned by an octree; for detals, please refer to [19]. Fgure 7 shows an eample Fgure 7. An eample of the probabltes P (D() = d k ), P 1(D() =d k ) and P bk accordng to the dstance d k. lne of sght D () = dk v p' φ p range mage Fgure 8. Our method fnds the nearest pont p for each patch. The standard devaton of d k s appromated by σ = σcosφ. of the probabltes P (D() =d k ), P 1 (D() =d k ) and P bk accordng to the dstance d k. The remanng ssue s how to compute P (D () d k ). As shown n Fgure 8, we fnd the nearest pont p for each patch of range mages from. A patch s usually a trangle that connects adjacent vertces. Though the pont s of the true surface ests along the lne of sght, snce the nearest pont p s nterpolated nsde of the patch, the dstrbuton of the nearest pont becomes smaller. In ths paper, we smplfy the computaton of p by appromatng the standard devaton of d k by σ = σcosφ. Therefore, P (D () d k ) s computed by P (D () d k )= 1 t (1 erf( 2 ) (16) 2σ where p =(1+t)(p ). 4. Eperments We evaluate our method usng synthetc data and real range mages. Frst, we compare our algnng and mergng methods wth prevous methods by usng synthetc range mages, to whch noses are added along the lne of sght. Second, we show that our method successfully creates a model of an object from nosy range mages. s 5

6 Table 1. Robustness of the estmated parameters. Our method Prevous method [15] X-std Z-std RMS Dr as vew drectons Fgure 9. Three eamples of slces of synthetc range mages n the z-plane. Ther shape and vew drectons are ndcated by colors. The orgnal mesh model s a thck black lne. The nterval of grds s Comparson by usng Synthetc Range Images We created one hundred synthetc range mages by addng nose of normal dstrbuton to a mesh model from ten dfferent vewponts. Fgure 9 shows an orgnal mesh model and three eamples of slces of synthetc range mages n the z-plane. The orgnal model conssts of two planes, whch are perpendcular to the z-plane. In ths case, the standard devaton of nose s 0.05 whle the wdth of a orgnal mesh model s 1.0 and the nterval between the vertces s The mamum dfference between vewng angles s 90 degrees Algnng Range Images We estmate the robustness of the algnment wth respect to the ntal parameters by comparng our method wth one of the prevous methods [15], whch fnds correspondng ponts along the algn of sght of one of range mages. Snce the orgnal mesh model conssts of two planes, t has ambguty f we algn the models. Thus, we estmate the error of the algnment results by comparng the dfference of translaton along the - and z-aes and the vew drecton from the ground truth. We algn two range mages startng wth varous ntal postons. The offsets from the ground truth are unformly dstrbuted wthn [ 0.25, 0.25] along each as. Table 1 shows the robustness of the estmated parameters. X-std. and Z-std. are the standard devatons of translaton along the - and z-aes. RMS s the root-mean-square dstance from the ground truth. Dr. s the dfference of the vewng drectons from the ground truth n degrees. Snce our method has smaller standard devaton and errors than [15], our method can estmate the parameters robustly wth respect to the varaton of the ntal parameters Mergng Range Images To test the proposed mergng algorthm, we merge two range mages that are acqured from dfferent vewponts. The postons of the nput range mages are at the ground truth. Fgure 10 shows two eamples of the results of mergng by the proposed method (red lne) and a prevous -as Fgure 10. The result of mergng 10 range mages. The red lne ndcates the results of the proposed method and the blue lne ndcates the results of the prevous method [19]. method [19] (blue lne), whch uses the average Eucldean dstance to the nearest ponts. To estmate the error of the merged mesh model, we compute the offset from the orgnal model along the z-as of the vertces. The RMS error of the nput range mages s We test ten tmes for each algorthm; the RMS error of the result of our method becomes , whle that of the prevous method s Our method obtans a better result than the prevous method and reduced the nose to 66% of the nput range mages Algnng and Mergng Real Range Images Net, we algn and merge real range mages captured by the Canesta DP200, whch obtans range mages at 30 frames/sec, but s less accurate than other slower laser range fnders. The standard devaton s 8mm along the lne of sght f an object s at about 1m from the sensor. We sequentally captured 72 range mages of an object shown n Fgure 11 from varous drectons by rotatng t on a turntable. Fgure 12 shows one of the range mages. From the top vew, we can see that the range mage contans large nose compared to the object sze, whch s about 420mm wde. We sequentally algned the range mages by parwse regstraton. The mages are algned automatcally by settng the ntal parameter of rotaton and translaton to be the same as the algned result of the prevous frame. Fgure 13 shows the estmated sensor poston and vewng drecton. Because the result of our method (red lne) becomes an accurate crcle, our method successfully algned the range mages whle the result of the prevous method [15] (blue lne) collapsed durng regstraton. 6

7 Fgure 11. A stuffed bear on a bo. z-as (mm) as (mm) (a) Front vew (b) Top vew Fgure 13. The estmated sensor poston and vewng drecton by our proposed method (red) and the prevous method [15] (blue). Fgure 12. A range mage of the stuffed bear contans large nose compared to the object sze. Fgure 15(a) shows the results of algnment, and Fgure 14(a) s a horzontal slce of the algned range mages. Though the shape of the object can be recognzed from them, t s represented as a cloud of ponts or meshes. By applyng our mergng method to the range mages, we construct a SDF. In Fgure 14(b) the red lnes are a slce of the mesh model etracted from the SDF. By comparng the merged model by the prevous method [19] (blue lnes), our method reduces the nose n the merged model. Fgure 15(b) s the result of mergng range mages. Snce our method does not assume any pror model of the object, the mesh model s not smooth. After smoothng the SDF by a mean flter, a smooth mesh model s etracted as shown n Fgure 15(c). 5. Concluson Ths paper proposed a probablstc method of algnng and mergng range mages. We formulated the ssues nvolved as mamum lkelhood estmates. By eamnng the error dstrbuton of a range fnder, we modeled t as a normal dstrbuton along the lne of sght. For algnng range mages our method estmates the parameters based on the EM approach. By assumng an error model, the algorthm s mplemented as an etenson of an ICP-based method. For mergng range mages, our method computes the sgned dstances by fndng the dstances of mamum lkelhood, and successfully reduced the sensor nose. Snce our proposed method uses multple correspondences for each verte of range mages, the errors after algnng and mergng range mages are less than those of earler methods that use one-to-one correspondences. In ths paper, we do not assume any pror model of the shape of a true surface. Thus, the resultng surface of mergng may not be smooth, as was (a) (b) Fgure 14. Horzontal slces of models: (a) algned range mages, (b) merged models by the prevous method [19] (blue) and the proposed method(red). shown n the results. Earler models have also been ntroduced durng the reconstructon of 3D models [27, 6]. In future work we ntend to ncorporate a pror model of the surface nto our framework to generate a smooth surface. References [1] P. Besl and N. McKay. A method for regstraton of 3-d shapes. IEEE Trans. Patt. Anal. Machne Intell., 14(2): , Feb [2] G. Blas and M. Levne. Regsterng multvew range data to create 3d computer objects. IEEE Transactons on Pattern Analyss and Machne Intellgence, 17(8): , [3] Canesta, Inc. CanestaVson EP Development Kt. [4] B. Curless and M. Levoy. A volumetrc method for buldng comple models from range mages. In Proc. SIG- GRAPH 96, pages ACM, [5] Cyra Technologes, Inc. Cyra [6] H. Dnh, G. Slabaugh, and G. Turk. Reconstructng surfaces usng ansotropc bass functons. In Proc. Internatonal Conference on Computer Vson, pages , Vancouver, Canada, July

8 (a) Algned range mages (b) Merged mesh model (c) Smoothed mesh model Fgure 15. The results of algnng and mergng range mages. (b) s etracted from the SDF and (c) s the result after smoothng the SDF. [7] J. Fredman, J. Bentley, and R. Fnkel. An algorthm for fndng best matches n logarthmc epected tme. ACM Transactons on Mathematcal Software, 3(3): , [8] H. Hoppe, T. DeRose, T. Duchamp, J. McDonald, and W. Stuetzle. Surface reconstructon from unorganzed ponts. In Proc. SIGGRAPH 92, pages ACM, [9] K. Ikeuch. Dgtally archvng cultural hertage. In Proc. 10th Internatonal Conference on Vrtual Systems and Multmeda (VSMM2004), November [10] A. Johnson and M. Hebert. Surface regstraton by matchng orented ponts. In Proc. Int. Conf. On Recent Advances n 3- D Dgtal Imagng and Modelng, pages , May [11] Konca Mnolta Photo Imagng U.S.A., Inc. Vvd 9 noncontact dgtzer. [12] M. Levoy, K. Pull, B. Curless, S. Rusnkewcz, D. Koller, L. Perera, M. Gnzton, S. Anderson, J. Davs, J. Gnsberg, J. Shade, and D. Fulk. The dgtal mchelangelo project: 3D scannng of large statues. In Proc. SIGGRAPH 2000, pages , [13] W. Lorensen and H. Clne. Marchng cubes: a hgh resoluton 3d surface constructon algorthm. In Proc. SIG- GRAPH 87, pages ACM, [14] P. Neugebauer. Geometrcal clonng of 3d objects va smultaneous regstraton of multple range mages. In Proc. Int. Conf. on Shape Modelng and Applcaton, pages , Mar [15] T. Osh, R. Sagawa, A. Nakazawa, R. Kurazume, and K. Ikeuch. Parallel algnment of a large number of range mages. In Proc. 3DIM 2003, pages , [16] I. Okatan and K. Deguch. A method for fne regstraton of multple vew range mages consderng the measurement error propertes. Computer Vson and Image Understandng, 87:66 77, [17] K. Pull. Multvew regstraton for large data sets. In Proc. Second Internatonal Conference on 3D Dgtal Imagng and Modelng, pages , [18] S. Rusnkewcz and M. Levoy. Effcent varant of the ICP algorthm. In Proceedngs of the 3rd Internatonal Conference on 3-D Dgtal Imagng and Modelng, pages , [19] R. Sagawa, K. Nshno, and K. Ikeuch. Adaptvely mergng large-scale range data wth reflectance propertes. IEEE Transactons on Pattern Analyss and Machne Intellgence, 27(3): , March [20] R. Sagawa, T. Osh, A. Nakazawa, R. Kurazume, and K. Ikeuch. Iteratve refnement of range mages wth ansotropc error dstrbuton. In Proc. IEEE/RSJ Internatonal Conference on Intellgent Robots and Systems, pages 79 85, October [21] R.Sagawa,N.Osawa,T.Echgo,andY.Yag. Realtme3d envronment modelng for a moble robot by algnng range mage sequences. In Proc. Brtsh Machne Vson Conference 2005, volume 1, pages , Oford, UK, September [22] K. Sato and S. Inokuch. Range-magng system utlzng nematc lqud crystal mask. In Proc. Internatonal Conference on Computer Vson, pages , [23] M. Soucy and D. Laurendeau. A general surface approach to the ntegraton of a set of range vews. IEEE Trans. Patt. Anal. Machne Intell., 17(4): , Aprl [24] F. Sten and G. Medon. Structural ndeng: effcent 3-d object recognton. IEEE Trans. Pattern Analyss and Machne Intellgence, 14(2): , [25] G. Turk and M. Levoy. Zppered polygon meshes from range mages. In Proc. SIGGRAPH 94, pages , Jul [26] M. D. Wheeler and K. Ikeuch. Sensor modelng, probablstc hypothess generaton, and robust localzaton for object recognton. IEEE Trans. Pattern Analyss and Machne Intellgence, 17(3): , March [27] R. Whtaker. A level-set approach to 3d reconstructon from range data. Internatonal Journal of Computer Vson, 29(3): , October [28] J. Wllams and M. Bennamoun. Multple vew surface regstraton wth error modelng and analyss. In Proc. IEEE Int. Conf. on Image Processng (ICIP 00), volume I, pages , Sep [29] Z. Zhang. Iteratve pont matchng for regstraton of freeform curves and surfaces. Internatonal Journal of Computer Vson, 13(2): ,

Range images. Range image registration. Examples of sampling patterns. Range images and range surfaces

Range images. Range image registration. Examples of sampling patterns. Range images and range surfaces Range mages For many structured lght scanners, the range data forms a hghly regular pattern known as a range mage. he samplng pattern s determned by the specfc scanner. Range mage regstraton 1 Examples

More information

A Range Image Refinement Technique for Multi-view 3D Model Reconstruction

A Range Image Refinement Technique for Multi-view 3D Model Reconstruction A Range Image Refnement Technque for Mult-vew 3D Model Reconstructon Soon-Yong Park and Mural Subbarao Electrcal and Computer Engneerng State Unversty of New York at Stony Brook, USA E-mal: parksy@ece.sunysb.edu

More information

Multi-stable Perception. Necker Cube

Multi-stable Perception. Necker Cube Mult-stable Percepton Necker Cube Spnnng dancer lluson, Nobuuk Kaahara Fttng and Algnment Computer Vson Szelsk 6.1 James Has Acknowledgment: Man sldes from Derek Hoem, Lana Lazebnk, and Grauman&Lebe 2008

More information

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,

More information

Structure from Motion

Structure from Motion Structure from Moton Structure from Moton For now, statc scene and movng camera Equvalentl, rgdl movng scene and statc camera Lmtng case of stereo wth man cameras Lmtng case of multvew camera calbraton

More information

A Volumetric Approach for Interactive 3D Modeling

A Volumetric Approach for Interactive 3D Modeling A Volumetrc Approach for Interactve 3D Modelng Dragan Tubć Patrck Hébert Computer Vson and Systems Laboratory Laval Unversty, Ste-Foy, Québec, Canada, G1K 7P4 Dens Laurendeau E-mal: (tdragan, hebert, laurendeau)@gel.ulaval.ca

More information

CS 534: Computer Vision Model Fitting

CS 534: Computer Vision Model Fitting CS 534: Computer Vson Model Fttng Sprng 004 Ahmed Elgammal Dept of Computer Scence CS 534 Model Fttng - 1 Outlnes Model fttng s mportant Least-squares fttng Maxmum lkelhood estmaton MAP estmaton Robust

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

A Robust Method for Estimating the Fundamental Matrix

A Robust Method for Estimating the Fundamental Matrix Proc. VIIth Dgtal Image Computng: Technques and Applcatons, Sun C., Talbot H., Ourseln S. and Adraansen T. (Eds.), 0- Dec. 003, Sydney A Robust Method for Estmatng the Fundamental Matrx C.L. Feng and Y.S.

More information

Efficient Reconstruction of Indoor Scenes with Color

Efficient Reconstruction of Indoor Scenes with Color Effcent Reconstructon of Indoor Scenes wth Color Ru Wang, Davd Luebke Department of Computer Scence Unversty of Vrgna {rw2p, luebke}@cs.vrgna.edu Abstract In ths paper we present an effcent and general

More information

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision SLAM Summer School 2006 Practcal 2: SLAM usng Monocular Vson Javer Cvera, Unversty of Zaragoza Andrew J. Davson, Imperal College London J.M.M Montel, Unversty of Zaragoza. josemar@unzar.es, jcvera@unzar.es,

More information

Machine Learning: Algorithms and Applications

Machine Learning: Algorithms and Applications 14/05/1 Machne Learnng: Algorthms and Applcatons Florano Zn Free Unversty of Bozen-Bolzano Faculty of Computer Scence Academc Year 011-01 Lecture 10: 14 May 01 Unsupervsed Learnng cont Sldes courtesy of

More information

Feature-Preserving Mesh Denoising via Bilateral Normal Filtering

Feature-Preserving Mesh Denoising via Bilateral Normal Filtering Feature-Preservng Mesh Denosng va Blateral Normal Flterng Ka-Wah Lee, Wen-Png Wang Computer Graphcs Group Department of Computer Scence, The Unversty of Hong Kong kwlee@cs.hku.hk, wenpng@cs.hku.hk Abstract

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

Accounting for the Use of Different Length Scale Factors in x, y and z Directions

Accounting for the Use of Different Length Scale Factors in x, y and z Directions 1 Accountng for the Use of Dfferent Length Scale Factors n x, y and z Drectons Taha Soch (taha.soch@kcl.ac.uk) Imagng Scences & Bomedcal Engneerng, Kng s College London, The Rayne Insttute, St Thomas Hosptal,

More information

Image Alignment CSC 767

Image Alignment CSC 767 Image Algnment CSC 767 Image algnment Image from http://graphcs.cs.cmu.edu/courses/15-463/2010_fall/ Image algnment: Applcatons Panorama sttchng Image algnment: Applcatons Recognton of object nstances

More information

Shape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram

Shape Representation Robust to the Sketching Order Using Distance Map and Direction Histogram Shape Representaton Robust to the Sketchng Order Usng Dstance Map and Drecton Hstogram Department of Computer Scence Yonse Unversty Kwon Yun CONTENTS Revew Topc Proposed Method System Overvew Sketch Normalzaton

More information

Simplification of 3D Meshes

Simplification of 3D Meshes Smplfcaton of 3D Meshes Addy Ngan /4/00 Outlne Motvaton Taxonomy of smplfcaton methods Hoppe et al, Mesh optmzaton Hoppe, Progressve meshes Smplfcaton of 3D Meshes 1 Motvaton Hgh detaled meshes becomng

More information

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton

More information

High-Boost Mesh Filtering for 3-D Shape Enhancement

High-Boost Mesh Filtering for 3-D Shape Enhancement Hgh-Boost Mesh Flterng for 3-D Shape Enhancement Hrokazu Yagou Λ Alexander Belyaev y Damng We z Λ y z ; ; Shape Modelng Laboratory, Unversty of Azu, Azu-Wakamatsu 965-8580 Japan y Computer Graphcs Group,

More information

An efficient method to build panoramic image mosaics

An efficient method to build panoramic image mosaics An effcent method to buld panoramc mage mosacs Pattern Recognton Letters vol. 4 003 Dae-Hyun Km Yong-In Yoon Jong-Soo Cho School of Electrcal Engneerng and Computer Scence Kyungpook Natonal Unv. Abstract

More information

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,

More information

CORRELATION ICP ALGORITHM FOR POSE ESTIMATION BASED ON LOCAL AND GLOBAL FEATURES

CORRELATION ICP ALGORITHM FOR POSE ESTIMATION BASED ON LOCAL AND GLOBAL FEATURES CORRELATION ICP ALGORITHM FOR POSE ESTIMATION BASED ON LOCAL AND GLOBAL FEATURES Marco A. Chavarra, Gerald Sommer Cogntve Systems Group. Chrstan-Albrechts-Unversty of Kel, D-2498 Kel, Germany {mc,gs}@ks.nformatk.un-kel.de

More information

The Research of Ellipse Parameter Fitting Algorithm of Ultrasonic Imaging Logging in the Casing Hole

The Research of Ellipse Parameter Fitting Algorithm of Ultrasonic Imaging Logging in the Casing Hole Appled Mathematcs, 04, 5, 37-3 Publshed Onlne May 04 n ScRes. http://www.scrp.org/journal/am http://dx.do.org/0.436/am.04.584 The Research of Ellpse Parameter Fttng Algorthm of Ultrasonc Imagng Loggng

More information

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng

More information

3D Virtual Eyeglass Frames Modeling from Multiple Camera Image Data Based on the GFFD Deformation Method

3D Virtual Eyeglass Frames Modeling from Multiple Camera Image Data Based on the GFFD Deformation Method NICOGRAPH Internatonal 2012, pp. 114-119 3D Vrtual Eyeglass Frames Modelng from Multple Camera Image Data Based on the GFFD Deformaton Method Norak Tamura, Somsangouane Sngthemphone and Katsuhro Ktama

More information

Distance Calculation from Single Optical Image

Distance Calculation from Single Optical Image 17 Internatonal Conference on Mathematcs, Modellng and Smulaton Technologes and Applcatons (MMSTA 17) ISBN: 978-1-6595-53-8 Dstance Calculaton from Sngle Optcal Image Xao-yng DUAN 1,, Yang-je WEI 1,,*

More information

X- Chart Using ANOM Approach

X- Chart Using ANOM Approach ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are

More information

Detection of an Object by using Principal Component Analysis

Detection of an Object by using Principal Component Analysis Detecton of an Object by usng Prncpal Component Analyss 1. G. Nagaven, 2. Dr. T. Sreenvasulu Reddy 1. M.Tech, Department of EEE, SVUCE, Trupath, Inda. 2. Assoc. Professor, Department of ECE, SVUCE, Trupath,

More information

3D vector computer graphics

3D vector computer graphics 3D vector computer graphcs Paolo Varagnolo: freelance engneer Padova Aprl 2016 Prvate Practce ----------------------------------- 1. Introducton Vector 3D model representaton n computer graphcs requres

More information

A NEW IMPLEMENTATION OF THE ICP ALGORITHM FOR 3D SURFACE REGISTRATION USING A COMPREHENSIVE LOOK UP MATRIX

A NEW IMPLEMENTATION OF THE ICP ALGORITHM FOR 3D SURFACE REGISTRATION USING A COMPREHENSIVE LOOK UP MATRIX A NEW IMPLEMENTATION OF THE ICP ALGORITHM FOR 3D SURFACE REGISTRATION USING A COMPREHENSIVE LOOK UP MATRIX A. Almhde, C. Léger, M. Derche 2 and R. Lédée Laboratory of Electroncs, Sgnals and Images (LESI),

More information

IMPLEMENTATION OF 3D SHAPE RECONSTRUCTION FROM RANGE IMAGES FOR OBJECT DIGITAL MODELING

IMPLEMENTATION OF 3D SHAPE RECONSTRUCTION FROM RANGE IMAGES FOR OBJECT DIGITAL MODELING IMPLEMENTATION OF 3D SHAPE RECONSTRUCTION FROM RANGE IMAGES FOR OBJECT DIGITAL MODELING Landecr A. Albuquerque Automaton and Control Group (GRACO), Department of Mechancal Engneerng, Faculty of Technology,

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

Support Vector Machines. CS534 - Machine Learning

Support Vector Machines. CS534 - Machine Learning Support Vector Machnes CS534 - Machne Learnng Perceptron Revsted: Lnear Separators Bnar classfcaton can be veed as the task of separatng classes n feature space: b > 0 b 0 b < 0 f() sgn( b) Lnear Separators

More information

SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE

SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE SHAPE RECOGNITION METHOD BASED ON THE k-nearest NEIGHBOR RULE Dorna Purcaru Faculty of Automaton, Computers and Electroncs Unersty of Craoa 13 Al. I. Cuza Street, Craoa RO-1100 ROMANIA E-mal: dpurcaru@electroncs.uc.ro

More information

SIMULTANEOUS REGISTRATION OF MULTIPLE VIEWS OF A 3D OBJECT

SIMULTANEOUS REGISTRATION OF MULTIPLE VIEWS OF A 3D OBJECT SIMULTANEOUS REGISTRATION OF MULTIPLE VIEWS OF A 3D OBJECT Helmut Pottmann a, Stefan Leopoldseder a, Mchael Hofer a a Insttute of Geometry, Venna Unversty of Technology, Wedner Hauptstr. 8 10, A 1040 Wen,

More information

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour 6.854 Advanced Algorthms Petar Maymounkov Problem Set 11 (November 23, 2005) Wth: Benjamn Rossman, Oren Wemann, and Pouya Kheradpour Problem 1. We reduce vertex cover to MAX-SAT wth weghts, such that the

More information

UAV global pose estimation by matching forward-looking aerial images with satellite images

UAV global pose estimation by matching forward-looking aerial images with satellite images The 2009 IEEE/RSJ Internatonal Conference on Intellgent Robots and Systems October -5, 2009 St. Lous, USA UAV global pose estmaton by matchng forward-lookng aeral mages wth satellte mages Kl-Ho Son, Youngbae

More information

A Novel Accurate Algorithm to Ellipse Fitting for Iris Boundary Using Most Iris Edges. Mohammad Reza Mohammadi 1, Abolghasem Raie 2

A Novel Accurate Algorithm to Ellipse Fitting for Iris Boundary Using Most Iris Edges. Mohammad Reza Mohammadi 1, Abolghasem Raie 2 A Novel Accurate Algorthm to Ellpse Fttng for Irs Boundar Usng Most Irs Edges Mohammad Reza Mohammad 1, Abolghasem Rae 2 1. Department of Electrcal Engneerng, Amrabr Unverst of Technolog, Iran. mrmohammad@aut.ac.r

More information

LECTURE : MANIFOLD LEARNING

LECTURE : MANIFOLD LEARNING LECTURE : MANIFOLD LEARNING Rta Osadchy Some sldes are due to L.Saul, V. C. Raykar, N. Verma Topcs PCA MDS IsoMap LLE EgenMaps Done! Dmensonalty Reducton Data representaton Inputs are real-valued vectors

More information

AUTOMATED REGISTRATION OF UNORGANISED POINT CLOUDS FROM TERRESTRIAL LASER SCANNERS

AUTOMATED REGISTRATION OF UNORGANISED POINT CLOUDS FROM TERRESTRIAL LASER SCANNERS AUTOMATED REGISTRATION OF UNORGANISED POINT CLOUDS FROM TERRESTRIAL LASER SCANNERS Kwang-Ho Bae and Derek D. Lcht Department of Spatal Scences, Curtn Unversty of Technology, GPO Box U1987, Perth, WA, 6845,

More information

Machine Learning. K-means Algorithm

Machine Learning. K-means Algorithm Macne Learnng CS 6375 --- Sprng 2015 Gaussan Mture Model GMM pectaton Mamzaton M Acknowledgement: some sldes adopted from Crstoper Bsop Vncent Ng. 1 K-means Algortm Specal case of M Goal: represent a data

More information

Semarang, Indonesia. Sepuluh Nopember Institute of Technology, Surabaya, Indonesia

Semarang, Indonesia. Sepuluh Nopember Institute of Technology, Surabaya, Indonesia 3D Surfaces Reconstructon of Seafloor Images Usng Multvew Camera Based on Image Regstraton Pulung Nurtanto Andono 1,a, I Ketut Eddy Purnama 2,b, Mochamad Harad 2,c, Tach Watanabe 3,d, and Kuno Kondo 3,e

More information

Reducing Frame Rate for Object Tracking

Reducing Frame Rate for Object Tracking Reducng Frame Rate for Object Trackng Pavel Korshunov 1 and We Tsang Oo 2 1 Natonal Unversty of Sngapore, Sngapore 11977, pavelkor@comp.nus.edu.sg 2 Natonal Unversty of Sngapore, Sngapore 11977, oowt@comp.nus.edu.sg

More information

RECONSTRUCTION OF 3D TUBULAR STRUCTURES FROM CONE-BEAM PROJECTIONS

RECONSTRUCTION OF 3D TUBULAR STRUCTURES FROM CONE-BEAM PROJECTIONS RECONSTRUCTION OF 3D TUBULAR STRUCTURES FROM CONE-BEAM PROJECTIONS Ja L Dept. of Elect. and Comp. Engrg. Oakland Unversty Rochester, MI 48309, USA Laurent Cohen CEREMADE Unversty Pars - Dauphne 75775 Pars

More information

MOTION PANORAMA CONSTRUCTION FROM STREAMING VIDEO FOR POWER- CONSTRAINED MOBILE MULTIMEDIA ENVIRONMENTS XUNYU PAN

MOTION PANORAMA CONSTRUCTION FROM STREAMING VIDEO FOR POWER- CONSTRAINED MOBILE MULTIMEDIA ENVIRONMENTS XUNYU PAN MOTION PANORAMA CONSTRUCTION FROM STREAMING VIDEO FOR POWER- CONSTRAINED MOBILE MULTIMEDIA ENVIRONMENTS by XUNYU PAN (Under the Drecton of Suchendra M. Bhandarkar) ABSTRACT In modern tmes, more and more

More information

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1 4/14/011 Outlne Dscrmnatve classfers for mage recognton Wednesday, Aprl 13 Krsten Grauman UT-Austn Last tme: wndow-based generc obect detecton basc ppelne face detecton wth boostng as case study Today:

More information

Real-time Motion Capture System Using One Video Camera Based on Color and Edge Distribution

Real-time Motion Capture System Using One Video Camera Based on Color and Edge Distribution Real-tme Moton Capture System Usng One Vdeo Camera Based on Color and Edge Dstrbuton YOSHIAKI AKAZAWA, YOSHIHIRO OKADA, AND KOICHI NIIJIMA Graduate School of Informaton Scence and Electrcal Engneerng,

More information

Line-based Camera Movement Estimation by Using Parallel Lines in Omnidirectional Video

Line-based Camera Movement Estimation by Using Parallel Lines in Omnidirectional Video 01 IEEE Internatonal Conference on Robotcs and Automaton RverCentre, Sant Paul, Mnnesota, USA May 14-18, 01 Lne-based Camera Movement Estmaton by Usng Parallel Lnes n Omndrectonal Vdeo Ryosuke kawansh,

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

Computer Animation and Visualisation. Lecture 4. Rigging / Skinning

Computer Animation and Visualisation. Lecture 4. Rigging / Skinning Computer Anmaton and Vsualsaton Lecture 4. Rggng / Sknnng Taku Komura Overvew Sknnng / Rggng Background knowledge Lnear Blendng How to decde weghts? Example-based Method Anatomcal models Sknnng Assume

More information

MOTION BLUR ESTIMATION AT CORNERS

MOTION BLUR ESTIMATION AT CORNERS Gacomo Boracch and Vncenzo Caglot Dpartmento d Elettronca e Informazone, Poltecnco d Mlano, Va Ponzo, 34/5-20133 MILANO boracch@elet.polm.t, caglot@elet.polm.t Keywords: Abstract: Pont Spread Functon Parameter

More information

Mathematics 256 a course in differential equations for engineering students

Mathematics 256 a course in differential equations for engineering students Mathematcs 56 a course n dfferental equatons for engneerng students Chapter 5. More effcent methods of numercal soluton Euler s method s qute neffcent. Because the error s essentally proportonal to the

More information

Real-time Joint Tracking of a Hand Manipulating an Object from RGB-D Input

Real-time Joint Tracking of a Hand Manipulating an Object from RGB-D Input Real-tme Jont Tracng of a Hand Manpulatng an Object from RGB-D Input Srnath Srdhar 1 Franzsa Mueller 1 Mchael Zollhöfer 1 Dan Casas 1 Antt Oulasvrta 2 Chrstan Theobalt 1 1 Max Planc Insttute for Informatcs

More information

Face Recognition University at Buffalo CSE666 Lecture Slides Resources:

Face Recognition University at Buffalo CSE666 Lecture Slides Resources: Face Recognton Unversty at Buffalo CSE666 Lecture Sldes Resources: http://www.face-rec.org/algorthms/ Overvew of face recognton algorthms Correlaton - Pxel based correspondence between two face mages Structural

More information

The Great Buddha Project: Modeling Cultural Heritage for VR Systems through Observation

The Great Buddha Project: Modeling Cultural Heritage for VR Systems through Observation The Great Buddha Project: Modelng Cultural Hertage for VR Systems through Observaton Katsush Ieuch, Atsush Naazawa, Kazuhde Hasegawa, Taesh Ohsh Insttute of Industral Scence, The Unversty of Toyo {, naazawa,

More information

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana

More information

Unsupervised Learning

Unsupervised Learning Pattern Recognton Lecture 8 Outlne Introducton Unsupervsed Learnng Parametrc VS Non-Parametrc Approach Mxture of Denstes Maxmum-Lkelhood Estmates Clusterng Prof. Danel Yeung School of Computer Scence and

More information

Image warping and stitching May 5 th, 2015

Image warping and stitching May 5 th, 2015 Image warpng and sttchng Ma 5 th, 2015 Yong Jae Lee UC Davs PS2 due net Frda Announcements 2 Last tme Interactve segmentaton Feature-based algnment 2D transformatons Affne ft RANSAC 3 1 Algnment problem

More information

Takahiro ISHIKAWA Takahiro Ishikawa Takahiro Ishikawa Takeo KANADE

Takahiro ISHIKAWA Takahiro Ishikawa Takahiro Ishikawa Takeo KANADE Takahro ISHIKAWA Takahro Ishkawa Takahro Ishkawa Takeo KANADE Monocular gaze estmaton s usually performed by locatng the pupls, and the nner and outer eye corners n the mage of the drver s head. Of these

More information

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 15

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 15 CS434a/541a: Pattern Recognton Prof. Olga Veksler Lecture 15 Today New Topc: Unsupervsed Learnng Supervsed vs. unsupervsed learnng Unsupervsed learnng Net Tme: parametrc unsupervsed learnng Today: nonparametrc

More information

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification

12/2/2009. Announcements. Parametric / Non-parametric. Case-Based Reasoning. Nearest-Neighbor on Images. Nearest-Neighbor Classification Introducton to Artfcal Intellgence V22.0472-001 Fall 2009 Lecture 24: Nearest-Neghbors & Support Vector Machnes Rob Fergus Dept of Computer Scence, Courant Insttute, NYU Sldes from Danel Yeung, John DeNero

More information

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes SPH3UW Unt 7.3 Sphercal Concave Mrrors Page 1 of 1 Notes Physcs Tool box Concave Mrror If the reflectng surface takes place on the nner surface of the sphercal shape so that the centre of the mrror bulges

More information

Efficiency of grid representation and its algorithms for areal 3D scan data

Efficiency of grid representation and its algorithms for areal 3D scan data Journal of Mechancal Scence and Technology 23 (2009) 1234~1243 Journal of Mechancal Scence and Technology www.sprngerlnk.com/content/1738-494x DOI 10.1007/s12206-008-1014-4 Effcency of grd representaton

More information

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach Angle Estmaton and Correcton of Hand Wrtten, Textual and Large areas of Non-Textual Document Images: A Novel Approach D.R.Ramesh Babu Pyush M Kumat Mahesh D Dhannawat PES Insttute of Technology Research

More information

Hermite Splines in Lie Groups as Products of Geodesics

Hermite Splines in Lie Groups as Products of Geodesics Hermte Splnes n Le Groups as Products of Geodescs Ethan Eade Updated May 28, 2017 1 Introducton 1.1 Goal Ths document defnes a curve n the Le group G parametrzed by tme and by structural parameters n the

More information

Mobile Robot Localization and Mapping by Scan Matching using Laser Reflection Intensity of the SOKUIKI Sensor

Mobile Robot Localization and Mapping by Scan Matching using Laser Reflection Intensity of the SOKUIKI Sensor Moble Robot Localzaton and Mappng by Scan Matchng usng Reflecton Intensty of the SOKUIKI Sensor *HARA Yoshtaka, KAWATA Hrohko, OHYA Akhsa, YUTA Shn ch Intellgent Robot Laboratory Unversty of Tsukuba 1-1-1,

More information

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Proceedngs of the Wnter Smulaton Conference M E Kuhl, N M Steger, F B Armstrong, and J A Jones, eds A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Mark W Brantley Chun-Hung

More information

Classifier Selection Based on Data Complexity Measures *

Classifier Selection Based on Data Complexity Measures * Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.

More information

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique //00 :0 AM Outlne and Readng The Greedy Method The Greedy Method Technque (secton.) Fractonal Knapsack Problem (secton..) Task Schedulng (secton..) Mnmum Spannng Trees (secton.) Change Money Problem Greedy

More information

Fitting & Matching. Lecture 4 Prof. Bregler. Slides from: S. Lazebnik, S. Seitz, M. Pollefeys, A. Effros.

Fitting & Matching. Lecture 4 Prof. Bregler. Slides from: S. Lazebnik, S. Seitz, M. Pollefeys, A. Effros. Fttng & Matchng Lecture 4 Prof. Bregler Sldes from: S. Lazebnk, S. Setz, M. Pollefeys, A. Effros. How do we buld panorama? We need to match (algn) mages Matchng wth Features Detect feature ponts n both

More information

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z.

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z. TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS Muradalyev AZ Azerbajan Scentfc-Research and Desgn-Prospectng Insttute of Energetc AZ1012, Ave HZardab-94 E-mal:aydn_murad@yahoocom Importance of

More information

Generalized-ICP. Aleksandr V. Segal Stanford University Dirk Haehnel. Sebastian Thrun.

Generalized-ICP. Aleksandr V. Segal Stanford University   Dirk Haehnel. Sebastian Thrun. Generalzed-ICP Aleksandr V. Segal Stanford Unversty Emal: avsegal@cs.stanford.edu Drk Haehnel Stanford Unversty Emal: haehnel@stanford.edu Sebastan hrun Stanford Unversty Emal: thrun@stanford.edu Abstract

More information

Multi-view 3D Position Estimation of Sports Players

Multi-view 3D Position Estimation of Sports Players Mult-vew 3D Poston Estmaton of Sports Players Robbe Vos and Wlle Brnk Appled Mathematcs Department of Mathematcal Scences Unversty of Stellenbosch, South Afrca Emal: vosrobbe@gmal.com Abstract The problem

More information

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 An Iteratve Soluton Approach to Process Plant Layout usng Mxed

More information

Positive Semi-definite Programming Localization in Wireless Sensor Networks

Positive Semi-definite Programming Localization in Wireless Sensor Networks Postve Sem-defnte Programmng Localzaton n Wreless Sensor etworks Shengdong Xe 1,, Jn Wang, Aqun Hu 1, Yunl Gu, Jang Xu, 1 School of Informaton Scence and Engneerng, Southeast Unversty, 10096, anjng Computer

More information

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data Malaysan Journal of Mathematcal Scences 11(S) Aprl : 35 46 (2017) Specal Issue: The 2nd Internatonal Conference and Workshop on Mathematcal Analyss (ICWOMA 2016) MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES

More information

RELATIVE ORIENTATION ESTIMATION OF VIDEO STREAMS FROM A SINGLE PAN-TILT-ZOOM CAMERA. Commission I, WG I/5

RELATIVE ORIENTATION ESTIMATION OF VIDEO STREAMS FROM A SINGLE PAN-TILT-ZOOM CAMERA. Commission I, WG I/5 RELATIVE ORIENTATION ESTIMATION OF VIDEO STREAMS FROM A SINGLE PAN-TILT-ZOOM CAMERA Taeyoon Lee a, *, Taeung Km a, Gunho Sohn b, James Elder a a Department of Geonformatc Engneerng, Inha Unersty, 253 Yonghyun-dong,

More information

Parallel matrix-vector multiplication

Parallel matrix-vector multiplication Appendx A Parallel matrx-vector multplcaton The reduced transton matrx of the three-dmensonal cage model for gel electrophoress, descrbed n secton 3.2, becomes excessvely large for polymer lengths more

More information

Edge Detection in Noisy Images Using the Support Vector Machines

Edge Detection in Noisy Images Using the Support Vector Machines Edge Detecton n Nosy Images Usng the Support Vector Machnes Hlaro Gómez-Moreno, Saturnno Maldonado-Bascón, Francsco López-Ferreras Sgnal Theory and Communcatons Department. Unversty of Alcalá Crta. Madrd-Barcelona

More information

Biostatistics 615/815

Biostatistics 615/815 The E-M Algorthm Bostatstcs 615/815 Lecture 17 Last Lecture: The Smplex Method General method for optmzaton Makes few assumptons about functon Crawls towards mnmum Some recommendatons Multple startng ponts

More information

A New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1

A New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1 A New Feature of Unformty of Image Texture Drectons Concdng wth the Human Eyes Percepton Xng-Jan He, De-Shuang Huang, Yue Zhang, Tat-Mng Lo 2, and Mchael R. Lyu 3 Intellgent Computng Lab, Insttute of Intellgent

More information

Multi-Resolution Geometric Fusion

Multi-Resolution Geometric Fusion Internatonal Conference on Recent Advances n 3-D Dgtal Imagng and Modellng, Ottawa, Canada May 12 15, 1997 Mult-Resoluton Geometrc Fuson Adran Hlton and John Illngworth Centre for Vson, Speech and Sgnal

More information

LEAST SQUARES. RANSAC. HOUGH TRANSFORM.

LEAST SQUARES. RANSAC. HOUGH TRANSFORM. LEAS SQUARES. RANSAC. HOUGH RANSFORM. he sldes are from several sources through James Has (Brown); Srnvasa Narasmhan (CMU); Slvo Savarese (U. of Mchgan); Bll Freeman and Antono orralba (MI), ncludng ther

More information

Computer Vision I. Xbox Kinnect: Rectification. The Fundamental matrix. Stereo III. CSE252A Lecture 16. Example: forward motion

Computer Vision I. Xbox Kinnect: Rectification. The Fundamental matrix. Stereo III. CSE252A Lecture 16. Example: forward motion Xbox Knnect: Stereo III Depth map http://www.youtube.com/watch?v=7qrnwoo-8a CSE5A Lecture 6 Projected pattern http://www.youtube.com/watch?v=ceep7x-z4wy The Fundamental matrx Rectfcaton The eppolar constrant

More information

Scan Conversion & Shading

Scan Conversion & Shading Scan Converson & Shadng Thomas Funkhouser Prnceton Unversty C0S 426, Fall 1999 3D Renderng Ppelne (for drect llumnaton) 3D Prmtves 3D Modelng Coordnates Modelng Transformaton 3D World Coordnates Lghtng

More information

Real Time 3D Environment Modeling for a Mobile Robot by Aligning Range Image Sequences

Real Time 3D Environment Modeling for a Mobile Robot by Aligning Range Image Sequences Real Time 3D Environment Modeling for a Mobile Robot by Aligning Range Image Sequences Ryusuke Sagawa, Nanaho Osawa, Tomio Echigo and Yasushi Yagi Institute of Scientific and Industrial Research, Osaka

More information

Ecient Computation of the Most Probable Motion from Fuzzy. Moshe Ben-Ezra Shmuel Peleg Michael Werman. The Hebrew University of Jerusalem

Ecient Computation of the Most Probable Motion from Fuzzy. Moshe Ben-Ezra Shmuel Peleg Michael Werman. The Hebrew University of Jerusalem Ecent Computaton of the Most Probable Moton from Fuzzy Correspondences Moshe Ben-Ezra Shmuel Peleg Mchael Werman Insttute of Computer Scence The Hebrew Unversty of Jerusalem 91904 Jerusalem, Israel Emal:

More information

A New Approach For the Ranking of Fuzzy Sets With Different Heights

A New Approach For the Ranking of Fuzzy Sets With Different Heights New pproach For the ankng of Fuzzy Sets Wth Dfferent Heghts Pushpnder Sngh School of Mathematcs Computer pplcatons Thapar Unversty, Patala-7 00 Inda pushpndersnl@gmalcom STCT ankng of fuzzy sets plays

More information

Learning Ensemble of Local PDM-based Regressions. Yen Le Computational Biomedicine Lab Advisor: Prof. Ioannis A. Kakadiaris

Learning Ensemble of Local PDM-based Regressions. Yen Le Computational Biomedicine Lab Advisor: Prof. Ioannis A. Kakadiaris Learnng Ensemble of Local PDM-based Regressons Yen Le Computatonal Bomedcne Lab Advsor: Prof. Ioanns A. Kakadars 1 Problem statement Fttng a statstcal shape model (PDM) for mage segmentaton Callosum segmentaton

More information

Scan Conversion & Shading

Scan Conversion & Shading 1 3D Renderng Ppelne (for drect llumnaton) 2 Scan Converson & Shadng Adam Fnkelsten Prnceton Unversty C0S 426, Fall 2001 3DPrmtves 3D Modelng Coordnates Modelng Transformaton 3D World Coordnates Lghtng

More information

Solving Route Planning Using Euler Path Transform

Solving Route Planning Using Euler Path Transform Solvng Route Plannng Usng Euler Path ransform Y-Chong Zeng Insttute of Informaton Scence Academa Snca awan ychongzeng@s.snca.edu.tw Abstract hs paper presents a method to solve route plannng problem n

More information

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements Module 3: Element Propertes Lecture : Lagrange and Serendpty Elements 5 In last lecture note, the nterpolaton functons are derved on the bass of assumed polynomal from Pascal s trangle for the fled varable.

More information

N-View Point Set Registration: A Comparison

N-View Point Set Registration: A Comparison N-Vew Pont Set Regstraton: A Comparson S. J. Cunnngton and A. J. Centre for Vson, Speech and Sgnal Processng Unversty of Surrey, Guldford, GU2 5XH, UK fa.stoddart,s.cunnngtong@ee.surrey.ac.uk Abstract

More information

Programming in Fortran 90 : 2017/2018

Programming in Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Exercse 1 : Evaluaton of functon dependng on nput Wrte a program who evaluate the functon f (x,y) for any two user specfed values

More information

Reading. 14. Subdivision curves. Recommended:

Reading. 14. Subdivision curves. Recommended: eadng ecommended: Stollntz, Deose, and Salesn. Wavelets for Computer Graphcs: heory and Applcatons, 996, secton 6.-6., A.5. 4. Subdvson curves Note: there s an error n Stollntz, et al., secton A.5. Equaton

More information