基于RPCA和SVM的快速运动目标检测算法  

Fast Moving Object Detection Based on RPCA and SVM

作　　者：李阳 华驰[1] 唐庭龙[2] LI Yang;HUA Chi;TANG Ting-long(School of Internet of Things Engineering,Jiangsu Vocational College of Information Technology,Wuxi 514153,China;School of Computer and Information,Three Gorges University,Yichang Hubei 443002,China)

机构地区：[1]江苏信息职业技术学院物联网工程学院,江苏无锡214153 [2]三峡大学计算机与信息学院,湖北宜昌443002

出　　处：《计算机仿真》2022年第11期463-466,495,共5页Computer Simulation

基　　金：江苏省高等学校基础科学(自然科学)研究项目资助(21KJB520006);江苏省高等职业教育教学融合生产集成平台建设项目(2019(26)号);江苏信息职业技术学院研究项目(10072020028(001))。

摘　　要：针对传统基于鲁棒主成分分析(Robust Principal Component Analysis,RPCA)的扩展模型在进行运动目标检测时速度较慢的问题,提出了一种基于RPCA和支持向量机(Support Vector Machine,SVM)的快速运动目标检测算法。算法先使用RPCA对图像序列进行分解得到稀疏前景,同时对图像序列进行超像素分割;然后分别提取前景所对应超像素区域的特征,并使用支持向量机进行训练得到模型。在进行运动目标检测时,只要使用已知模型对超像素的特征进行判断操作便可知此超像素是否属于前景。实验结果表明,所提方法在保证F-measure值较高的同时,检测的速度得到大幅提升,是已有RPCA扩展模型的50倍左右。Aiming at the problem that the traditional extended robust principal component analysis(RPCA)model is slow in moving object detection,this paper proposes a fast moving object detection algorithm based on RPCA and support vector machine(SVM).RPCA was applied to decompose the image sequence to get sparse foreground.At the same time,superpixel segmentation was applied to image sequences.Then,the features of the superpixel region corresponding to the foreground and background were extracted respectively,and the model was obtained by training with SVM.In moving object detection,we only need to use the known model to judge the features of the superpixel,and then we can know whether the superpixel belongs to the foreground.The experimental results show that the proposed method not only guarantees a high F-measure value,but also greatly improves the detection speed,which is about 50 times of the existing extended RPCA model.

关 键 词：鲁棒主成分分析 运动目标检测 超像素 支持向量机 

分 类 号：TN391.41[电子电信—物理电子学]