论文部分内容阅读
针对目前基于速度检测公共场所密集人群异常行为存在的检测准确率低、使用范围局限的问题,从人群的加速度角度对可能导致公共安全事故的人群异常行为进行研究,提出了一种基于加速度检测人群异常行为的算法,并基于该算法实现了针对人群逃散、人群聚集、人群拥挤和人群逆行4种异常行为检测的系统。首先,利用金字塔Lucas-Kanade光流法进行特征点跟踪;然后,在获取到特征点的速度矩阵基础上进一步计算其加速度矩阵,反映速度的整体变化;最后,从加速度大小和方向两方面检测人群异常行为。结果表明,所提算法检测用时较少,相比基于速度检测的对比算法,检测的正确率提高到80%,误报率降低为5%。
Aiming at the current problem of low detection accuracy and limited scope of use of speed detection of abnormal crowd behavior in public places, this paper studies the abnormal behavior of people who may cause public safety accidents from the perspective of crowd acceleration, and proposes a new method based on acceleration detection Abnormal behavior algorithm, and based on the algorithm to achieve the escape of the crowd, the crowd gathered, crowded and retrograde retrograde retrograde behavior of four kinds of abnormal behavior of the system. Firstly, the feature points are tracked by the pyramid Lucas-Kanade optical flow method. Then, the acceleration matrix is further calculated based on the velocity matrix of the feature points to reflect the overall change of velocity. Finally, the population is detected from the magnitude and direction of acceleration Abnormal behavior. The results show that the proposed algorithm has less detection time, and the accuracy of detection is improved to 80% and the false alarm rate is reduced to 5% compared with the comparison algorithm based on speed detection.