ebook img

Moving object detection apparatus and moving object detection method PDF

72 Pages·2013·5.32 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Moving object detection apparatus and moving object detection method

US008599252B2 (12) United States Patent (10) Patent N0.: US 8,599,252 B2 Komoto et a1. (45) Date of Patent: Dec. 3, 2013 (54) MOVING OBJECT DETECTION APPARATUS FOREIGN PATENT DOCUMENTS AND MOVING OBJECT DETECTION METHOD JP 8-214289 8/1996 JP 2006-31114 2/2006 (75) Inventors: Ayako Komoto, Osaka (JP); Kunio JP 2008-146185 6/2008 Nobori, Osaka (JP); Masahiro IWasaki, WO 2010/050110 5/2010 Kyoto (JP) WO 2010/079556 7/2010 (73) Assignee: Panasonic Corporation, Osaka (JP) OTHER PUBLICATIONS (*) Notice: Subject to any disclaimer, the term of this patent is extended or adjusted under 35 International Search Report issued Nov. 1, 2011 in corresponding U.S.C. 154(b) by 86 days. International Application No. PCT/JP201 1/004173. P. Anandan, “A Computational Framework and an Algorithm for the (21) Appl. No.: 13/454,577 Measurement of Visual Motion”, International Journal of Computer Vision, vol. 2, Jan. 1989, pp. 283-310. (22) Filed: Apr. 24, 2012 (Continued) (65) Prior Publication Data US 2012/0206597 A1 Aug. 16,2012 Primary Examiner * Dave CZekaj Assistant Examiner * Tsion B Owens Related US. Application Data (74) Attorney, Agent, or Firm * Wenderoth, Lind & Ponack, (63) Continuation of application No. PCT/J P201 1/ 004173, L.L.P. ?led on Jul. 25, 2011. (30) Foreign Application Priority Data (57) ABSTRACT Jul. 27, 2010 (JP) ............................... .. 2010-168704 A moving object detection apparatus includes: a stationary measure calculation unit calculating, for each of trajectories, (51) Int. Cl. a stationary measure representing likelihood that the traj ec H04N 7/18 (2006.01) tory belongs to a stationary obj ect; a distance calculation unit G06K 9/00 (2006.01) calculating a distance representing similarity between traj ec G06K 9/34 (2006.01) tories; and a region detection unit (i) performing a transfor (52) US. Cl. mation based on the stationary measures and the distances USPC .......................... .. 348/135; 382/103; 382/173 between the trajectories, so that a ratio of a distance between (58) Field of Classi?cation Search a trajectory on stationary object and a trajectory on moving USPC ................................. .. 348/135; 382/103, 173 object, to a distance between trajectories both belonging to See application ?le for complete search history. stationary object becomes greater than a ratio obtained before the transformation and (ii) detecting the moving object region (56) References Cited by separating the trajectory on the moving object from the trajectory on the stationary object, based on a geodesic dis U.S. PATENT DOCUMENTS tance between the trajectories. 2011/0002509 A1 1/2011 Nobori et al. 2011/0013840 A1* 1/2011 Iwasaki et al. .............. .. 382/173 2011/0228987 A1 9/2011 Iwasaki et al. 32 Claims, 32 Drawing Sheets Camera I 110 .................. "4.1.991"... iMoving object detection :apparatus 101 ; Image receiving unit l 102 ' I Trajectory calculation unit j 103 104 \L 105 Stationary measure Distance Subclass calculation unit calculation unit classi?cation unit J 106 i I Weighted distance calculation unit I 107 Segmentation unit Display I 120 US 8,599,252 B2 Page 2 (56) References Cited Edsger W. Dijkstra, “A Note on Two Problems in Connexion with Graphs”, Numerische Mathematik, vol. 1, Dec. 1959, pp. 269-271. Pedro F. FelZensZwalb et al., “Ef?cient Graph-Based Image Segmen OTHER PUBLICATIONS tation”, International Journal of Computer Vision, vol. 59, No. 2, 2004, pp. 1-26. Vladimir Kolmogorov et al., “Computing Visual Correspondence Carlo Tomasi et al. “Detection and Tracking of Point Features”, KLT: with Occlusions via Graph Cuts”, International Conference on Com An Implementation of the Kanade-Lucas-Tomasi Feature Tracker, puter Vision, Mar. 7, 2001, pp. 1-37. http://www.ces.clemson.edu/"stb/klt/, version 1.3.2, Mar. 28, 2006, pp. 1-20. Jianbo Shi et al., “Good Features to Track”, IEEE Conference on Kenichi Kanatani, GaZo rikaii3 jigen ninshiki no suuri4(Image Computer Vision and Pattern Recognition, Jun. 1994, pp. 1-8. UnderstandingiMathematical Principle of Three-dimensional Rec Richard Hartley et al., “Multiple View Geometry in Computer ognition), Morikita Publishing Co., Ltd., May 24, 1990, pp. 78-99 Vision”, Second Edition, Cambridge University Press, 2003, pp. (with partial English translation). 87-131, 239-261, 279-309 and 365-407. Richard Hartley et al., “Multiple View Geometry in Computer Chang Yuan et al., “Detecting Motion Regions in the Presence of a Vision”, Second Edition, Cambridge University Press, 2003, pp. Strong Parallax from a Moving Camera by Multiview Geometric 87-131, 239-261, 279-309, 365-389 and 391-407. Constraints”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 29, No. 9, Sep. 2007, pp. 1627-1641. * cited by examiner US. Patent Dec. 3, 2013 Sheet 1 0f 32 US 8,599,252 B2 FIG. 1 110 Camera / /10() EMaopvpianrgat uosbj ect detection \ /10]_ ; 1 Image receiving unit I 5 y /102 5 i Trajectory calculation unit i \ 103 v 104 v 105 I ;I cSatlactuiloantairoyn muneiats ure cDailsctualnactei on unit Scluabscsliafiscsa tion unit : \ \ \ r/1O6 E Ii Weighted distanc\e calculatio/n1 u0ni7t E i Segmenta\ tion unit /12o Display FIG. 2 200 I. _____________________________ ..... -.| I Computer: /110 /201 202 Camera > I/F >< CPU r4 : 203 i ; 206 < > ROM ’‘ : /120 : Vd/ : D- | / : l e0 , 204 ; ‘Sp ay ; carcl ’ \ RAM K a A I l I \ V/ZOS I HDD ' e _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ . . _ _ - _ - _ - - _ - - - _ _ _ _ -_ US. Patent Dec. 3, 2013 Sheet 2 0f 32 US 8,599,252 B2 FIG. 3 Receive image data on a plurality (T number) of consecutive images i, 5302 Calculate motion information on pixels using temporally different images to generate trajectories over the images V 5303 v 5304 \ /S3O5 Classify the Calculate a stationary Cakulate a measure indicating distance trajectories llkEIlhODCl of belonglng representing Into subclasses which to a stationary object, Simi'arity between for.each .Of the the trajectories are subsets of trajectones the trajectories \y ,/S306 Calculate a geodesic distance based on the stationary measures of the trajectories, the distance between the trajectories, and the subclasses \/ S307 Output a result of segmentation obtained by separating or integrating the subclasses based on the geodesic distance between the subclasses US. Patent Dec. 3, 2013 Sheet 3 0f 32 US 8,599,252 B2 FIG. 5A L) UH FIG. 5B O UU FIG. 5C MI\ /Wl US. Patent Dec. 3, 2013 Sheet 4 0f 32 US 8,599,252 B2 FIG. 6A Pixel i 603a Input picture 601 Pixel k 603i) Motion vector information 602 t t+n FIG. 65 T number of pictures 604 included in input video sequence Trajectory Pixel i 605a xi 606a Pixel k 505i) Trajectory xk 606D 1 2 T ' 1 T (Frame) FIG. 7 Stationary measure / 103 calculation unit Geometric constraint /_’7O1 estimation unit \ Error calculation unit US. Patent Dec. 3, 2013 Sheet 5 0f 32 US 8,599,252 B2 FIG. 8 Stereo Trinocular Epipolar Triiinear constraint constraint Homography Structure constraint consistency constraint FIG. 9A Stationary point X Second frame First frame Point {)1 in image ‘:> Line l2 in image of of first frame second frame: epipoiar line FIG. 9B US. Patent iDec.3,2013 Sheet60f32 US 8,599,252 B2 FIG. 10 ®TT \ 1101 1102 zgEk ‘///T Ei FIG. 115 T T T E Smal|er<-———> Greater US. Patent Dec. 3, 2013 Sheet 7 0f 32 US 8,599,252 B2 FIG. 12 Distance calculation unit ’’ 104 Euclidean distance / 1201 calculation unit V Geodesic distance ’ 1202 calculation unit FIG. 13A I Trajectory i Trajectory s Trajectory j FIG. 135 Data pointJ' Data point s I US. Patent Dec. 3, 2013 Sheet 8 0f 32 US 8,599,252 B2 FIG. 13C Data point 5 Data pointj j,,iégééééigéiééé?égi :90, J') 1302 ' FIG. 14 Euclidean distance 1401 Geodesic distance 1402 Trajectory xi / I i Trajectory xS (a) Euclidean distance between (D) Geodesic distance between trajectories trajectories

Description:
348/135; 382/103; 382/173. Field of 107. Segmentation unit. 120. Display I . Output a result of segmentation obtained by separating or integrating
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.