Object Tracking by Using Optical Flows and Edges
This paper describes real time object tracking in a complex background by extracting optical flows, and integration of optical flows and contrast. In real time tracking, optical flows are calculated based on the generalized gradient model. The system is a linear array processor capable of processing directly TV images in real time. Our system detects a moving object, and tracks it from the optical flow data. The system controls a camera head to follow the object. Although the background moves, the movement of the object is discriminated from the background.
If two objects should be tracked simultaneously, each object is tracked simultaneously. When two objects overlap, foreground object is recognized, and each object is tracked without confusion.
If an object stops or overlaps with another object with a similar motion, edged extracted from an intensity image is used together with optical flows. While the object is moving, the contour is extracted using optical flows and the contrast. When the object cannot be tracked only with optical flows, the edge information is used to determine the the object contour.
KeywordsOptical Flow Null Point Foreground Object Flow Vector Object Region
Unable to display preview. Download preview PDF.
- Dickmanns, E.D., Behringer, R., Brudigam, C., Dickmanns, D., Thomanek, F., and Holt, V. v.: An All-Transputer Visual AutobahnAutopilot/Copilot, Proc. IEEE ICCV’93, pp. 608–615 (1993)Google Scholar
- Brandenbusch, M., Sung, Chen-Ko, and Wirtkopp, T.: A vision system for real-time motion analysis in image sequences, FhG IITB Mitteilungen pp. 31–35 (1992)Google Scholar
- Murray, D.W. and McLauchlan, and P.F. Heid, I.D. and Sharkey, P.M.: Reactions to Peripheral Image Motion using a Head/Eye Platform, Proc. ICCV’93, pp. 403–411 (1993)Google Scholar
- Inoue, H. and Tachikawa, T. and Inaba, M.: Robot Vision System with a Correlation Clip for Real-time Tracking, Optical Flow and Depth Map Generation, Proc. IEEE Int. Conf. RandA, Vol. 2, pp. 1621–1626 (1992)Google Scholar
- Sobey, P. and Srinivasan, M.V.: Measurement of optical flow using a generalized gradient scheme, Journal of the Optical Society of America, pp. 1488–1498 (1991).Google Scholar
- Chen, H. -J. and Shirai, Y.: Detecting Multiple Image Motions by Exploiting Temporal Coherence of Apparent Motion, Proc.IEEE Conf. on Computer Vision and Pattern Recognition, pp. 899–902 (1994)Google Scholar
- Etoh, M. and Shirai, Y.: Segmentation and 2D motion estimate by region fragments, Proc. 4th Int. Conf. Comput. Vision, pp. 192–199 (1993)Google Scholar
- Thompson, W.B.: Combining Motion and Contrast for Segmentation, IEEE Trans. Pattern Anal. Mach. Intell., Vol.PAMI—2, No. 6 pp. 543–549 (1980)Google Scholar
- Mae, Y. Yamamoto, S. Shirai, Y. and Miura, J.: Optical Flow Based Object Tracking by Active Camera, Proc. 2nd Japan- France Congress on Mechatronics, pp. 545–548 (1994).Google Scholar
- Yamamoto, S., Mae, Y., Shirai Y. and Miura, J.: Realtime Multiple Object Tracking Based on Optical Flows, Proc. IEEE International Conference on Robotics and Automation, pp. 2328–2333 (1995).Google Scholar