Moving Object Detection And Tracking By Scanning Monitoring In Camp's Area

In order to solve the problem of moving object auto-detection and auto-tracking of large field in camp's area, this paper proposed a new method which based on scanning monitoring.1. Background modeling is based on a Pan-Tilt-Zoom (PTZ) camera to scan in a certain cycle time without moving objects, and we get the vertical projection using the result of vertical edge detection. As processing real-time video, the most similar one to real-time video is found, by comparing the similarity between features map of real-time frame and background frame. The moving objects are extracted with background subtraction algorithm. Backgrounds are updated adaptively by a new method.2. A new coarse-to-fine background-matching algorithm is proposed to solve the failure of coarse matching when targets occupy a large proportion of the total image. SIFT is simplified to achieve background precise matching. The results from SIFT are also used to improve the transform parameters obtained by template matching3. Background is compensated using global compensation parameters which are gotten from real-time image and background image by template matching. Under the Close-range surveillance condition, the background deformation will easily lead to false target, so a flexible method of registration is proposed to secondary registrate on the target area. It can effectively remove the false target, enhance the integrity of the true goal, and improve the information Noise ratio.4. Based on the binary result, the image is devided into blocks. Adjacent regions were combined and foreground regions were located. We propose a new algorithm to track object using a covariance descriptor. We represent an object window as the covariance matrix of features and their correlation within the same representation. Then to search the whole image to find the region which has the smallest distance from the current object model. The best matching region determines the location of the object in the current frame to achieve the goal of tracking. The covariance matrix enables efficiently fusion of different types of features and modalities, and its dimensionality is small. The proposed algorithm was effective and efficient.Experiments show that the moving object auto-detection and auto-tracking algorithm base on scanning monitoring is efficient and suitable for the large field monitoring in camp's area.