Research On Abnormal Crowd Escape Beheavior Detection In Video Sequences

Intelligent video surveillance has been attracted more and more attention of researchers, mainly because it can solve security problems in public places.Intelligent video surveillance can automatically analyze population unusual event recognition scene, and timely warning, with a wide range of high value and social value. The main work is as follows:First, pre-processing using the filtering method. Mean filtering method is utilized to deal with the noise and the regular grid is used to divide the video sequences into a set of patches. The statistical averaging value of motion vectors is calculated in each patch to reduce the effect of noise and reduce the subsequent computation complexity, and the size of the patch is determined by the experiment. In this paper, the k-means algorithm is employed to extract the foreground area.Secondly, the improved acceleration feature is extracted to detect the anomalous crowd behaviors in video surveillance systems. Using a three-frame image brightness invariant after obtaining acceleration characteristics, combined with mass distribution index acceleration formula changes, taking into account when an emergency occurs, the population as a whole form will suddenly appear as scattered individuals and populations become too large acceleration.Therefore, we can use to detect changes in the distribution of population-based improvements acceleration characteristics. Experimental tests carried out by public databases, and abnormal behavior detection algorithm based on population and social forces carried out comparative experiments.Finally, the position of the crowd to locate abnormalities may occur. In this paper, to represent the position of the crowd exception may occur by introducing flee center. Abnormal localization algorithm based on a single escape center on using the method for improving the acceleration vector KNNS reverse extendedintersection set to be divided, and the physical center of the intersection area is densely flee center. In fact often the case of multiple simultaneous escape from the center, aiming to improve this situation by using a single algorithm and fled from the combination of a central dividing method based on the multi-centered escape abnormal localization method. By design and synthesis of data verify the theoretical feasibility, and the abnormal database UMN groups to assess the performance of the algorithm.