Research On Moving Target Detection Algorithm Based On Video

With the rapid development of computer technology and image processing technology in recent years,target detection system with moving object in video has attracted considerable attention. At different levels of industrial chains in social production, the technology of moving target detection has been developed and applied widely. Because of the broad applications in the market, more and more researchers have spent great amount of time on the target detection systems and proposed a variety of effective methods. Now the detection of moving objects has become one of the hot research topics for application prospects, which also has potential economic value in the field of intelligence surveillance.Moving target detection is a very important issue in the study of video and image, which refers to the region of interest being separated from the background. In this paper, moving target detection problems based on video surveillance, shadow detection and fall detection are researched. At the same time, analysis and solving methods are put forward in this paper. Compared with the previous methods, it is proved that the algorithms proposed in this paper are more suprior.The main contribution of this paper is as follows:1) A moving target detection algorithm based on developed frame difference and rough entropy is proposed. In the view of the shortcomings of time domain segmentation and spatial segmentation, this algorithm takes both the feature of time domain and space domain into consideration. Make regional division in space domain to obtain the spatial map according to rough entropy and gain the motion information with the improved frame difference. Then divide all areas of the target region into three different parts, including the target region with more moving target pixels, background region with less moving target pixels, and tentative region which will be finally classfied by the space-time region energy model. In conclusion, this algorithm could achieve better segmentation of foreground objects.2) A moving cast shadow detection method based on multiple features fusion is proposed in this paper. Firstly, three kinds of features such as intensity, color and texture are extracted by different measures from the foreground image. Then, the synthetic feature map is created by combination of these features with different weight. Consequently, moving cast shadow pixels are detected from their moving objects roughly. Finally, spatial adjustment is applied to correct misclassified pixels for acquiring the correct shadow detection result. The advantage of the proposed method is tested on various scenes. The results show that the method can obtain high detection rate.3) A fall detection algorithm based on contour tracing is proposed. Regarding the high misdiagnosis rate and complexity in traditional fall detection algorithms, the detection algorithm will first extract the old man occuring in the surveillance video. Then obtain the complete old man body shape with median filtering and morphological processing, and generate the boundary chain code according to contour tracking algorithm. Finally, whether the old man has fell down or not will be determined on the Euclidean distance between the actual boundary chain code and the template chain code. Experiments show that the novel algorithm could not only reduce computation complexity, but also significantly improve accuracy rate.