Research On Human Detecting Based On Infrared Images

With the development of modern socio-economic and the accelerated process of urbanization, the population increases quickly especially in some public places and it can be very dangerous in some situations. So the behavior of people crowd has drawn increasing attention. Crowd surveillance methods based on image processing have many advantages such as save manpower costs etc. Therefore surveillance of the population is of great realistic significance and application prospects.Because the crowd behavior is very complicated, there is no good method for intelligent control of the crowd. It can be very difficult for image processing because of overlap and complex texture of environment. There is a great difficulty for ordinary visible light camera and because the infrared image has some advantages in the image processing for people crowd, so we have tried to do some work on the infrared image processing for people crowd surveillance.Infrared camera has not used for people crowd surveillance yet, after the analysis of the infrared image on the basis of imaging mechanism and reading a lot of papers, this paper presents a general method for the surveillance of people crowd. Firstly infrared camera is selected and then some scenes are designed for dangerous situation and an infrared image library is build. After that some imaging processing algorithms are used for these IR images include motion estimation algorithms and density estimation algorithms. Motion estimation algorithms use differential optical flow method and block matching method. Some problems have been found in low view images which are caused by the near far effect, so some special experiments are designed to improve this problem and the result of the experiments show that this method has a good performance on low view images; Density estimation mainly uses character of people crowd to estimate number of people, these character include foreground area, background area and edge pixel numbers. The result of the experiment shows that this method is effective.Infrared images have great advantages especially for image segmentation, complex texture and dark environment compared with visible light images and are appropriate for those environments that have distinct temperature contrast. The ordinary image processing method used for visible light image can be used for infrared image. This paper have tried some algorithm for people crowd image processing using infrared image, and the result shows that infrared image do have some advantages compared to visible light image and have a great prospect.