摘要：Detecting and tracking up to 8-9 targets is already an arduous task for traditional methods. So it is even more challenging to deal with crowd objects. In this paper, segmentation and density analysis for video crowd flow is realized by a chaotic dynamics based approach. Crowd motion system is treated as a chaotic dynamic system, and crowd targets are treated as particles. As a result, Finite Time Lyapunov Exponent (FTLE) field is obtained. After morphological processing, motion region extraction and classification, texture information can be got from Gray Level Co-occurrence Matrix (GLCM). In GLCM analysis, density information can be achieved. Experimental results show the effectiveness of the proposed approach.
