结合复小波域去噪和PSO-TSVM的群体异常行为检测  

Detection of abnormal group activity by combining complex wavelet domain based denoising with PSO-TSVM

作　　者：胡根生[1,2] 吴玉林 梁栋 HU Gensheng;WU Yulin;LIANG Dong(School of Electronics and Information Engineering,Anhui University,Hefei 230601,China;Anhui Provincial Key Laboratory of Polarization Imaging Detection Technology,Hefei 230031,China)

机构地区：[1]安徽大学电子信息工程学院,安徽合肥230601 [2]偏振光成像探测技术安徽省重点实验室,安徽合肥230031

出　　处：《传感器与微系统》2020年第5期143-147,共5页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(61672032);偏振光成像探测技术安徽省重点实验室开放项目(2016-KFKT-003)。

摘　　要：为了提高群体异常行为检测准确率,减少异常检测中噪声带来的影响,给出一种结合复小波域去噪和粒子群优化孪生支持向量机(PSO-TSVM)的群体异常行为检测算法。通过Horn-Schunck光流法提取视频中群体行为的速度、加速度、方向特征和人群密度特征;利用非抽样对偶树复小波包变换和双变量模型对抽取的群体行为特征进行噪声去除;使用去噪后的群体行为特征训练和测试经粒子群算法优化的孪生支持向量机模型,实现视频中的群体异常行为检测。在UMN视频数据集和自建数据集上的实验结果表明:相较于社会力模型和粒子熵模型等方法,所提算法具有更高的检测准确率。In order to improve the detection accuracy of abnormal group activity and reduce the impact of noise in anomaly detection,an algorithm for the detection of abnormal group activity by combining complex wavelet domain based denoising with particle swarm optimization twin support vector machine(PSO-TSVM)is proposed.The Horn-Schunck optical flow method is used to extract the features of velocity,acceleration,direction of group activity and the feature of crowd density in the video.Non-subsampled dual-tree complex wavelet packet transform and bivariate model are used to remove the noises of the group activity features extracted from the video.Abnormal group activity in the video is detected by PSO-TSVM which is trained and tested using these denoised group activity features.Experimental results on UMN video dataset and self-built dataset show that the proposed algorithm has higher detection accuracy than the social force model and particle entropy model.

关 键 词：群体异常行为检测 非抽样对偶树复小波包变换 双变量模型 粒子群优化-孪生支持向量机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]