点击次数：

发布时间：2019-03-10

论文类型：会议论文

发表时间：2016-01-01

收录刊物：CPCI-S

关键字：Crowd Density; Prior Knowledge; Perspective correction; Exercise Intensity

摘要：In recent years, With the development of science and technology to promote the popularity of video surveillance, computer techniques have a great value on obtaining the crowd counting information of surveillance video automatically, but perspective effects, mutual occlusion between people and other factors make crowd counting difficult. This paper presents a crowd counting method on sparse scene. Firstly, analyzing the characteristics of the surveillance video to access available prior knowledge; Secondly, combining with prior knowledge to extract the characteristics of target prospects block; Finally, support vector regression machine is employed to estimate the number. Experiments show that the method improves the situation of pedestrians occlusion crowd counting estimation accuracy.