| In recent years, accidents frequent happened because of crowd congestion in stations, railway docks, squares scenic and other public places. While these public spaces are equipped with surveillance cameras, but existing monitoring tools mainly rely on manual monitoring, there must be staff to watch the video; this method can not either quickly and efficiently count the number of people in public places for accurate statistics, or make early warning for crowd congestion. This paper uses image processing technology, by analyzing surveillance video, proposes a new people counting method based on analysis of video, in order to gain the state of public places in real time, issue a warning to avoid unnecessary incidents depending on the situation. The main work of this paper is as follows:(1)Improve Gaussian mixture background model. The improved background modeling is low computational complexity, more concise expression, and stronger operability. Without compromising the quality of the background, the improved method reduces the computational complexity, detect target and update background more quickly.(2)Use background subtraction method to detect moving objects quickly and effectively in the video image(such as people, vehicles, etc.), analyze the histogram characteristics, use Shen operator to find an adaptive binarization threshold to complete the moving target zone detection.(3) Through the analysis of human skin color area and spatial characteristics of the non-skin areas, adopt classification techniques to classify these two areas, establish relevant color model, combine with the human body geometry, achieve accurate positioning of face, calculate the number of people according to face number.(4) Develop a system on the MTLAB platform based on the detection algorithm. Use surveillance video from the actual scene for experiment, the experimental results show that the proposed method can effectively detect the face region, and calculate the number of people according to the number of face regions. |
