Study On Video Moving Objects Detection And Real-Time Processing System

In the digital video signal processing the technology of video moving object detection is very important. It is widely used in military and industry. Along with the fast development of microelectronic technology, especially, the scale and speed of DSP and the processing technology innovation of digital video signal, the real-time implementation of video moving object detection turns into possibility, and the related theories, techniques and technologies becomes the dominating research field.The main contribution of the dissertation is the whole studying of the real-time detection technologies of video moving objects. In this dissertation the theoretical model for the detection of video moving objects has been set up, which offers the mathematic description about the video moving objects.When the detection system of video moving objects is tracking the moving objects under the servo control the camera is also moving, which results in the background movement. In order to segment the moving objects from moving video background it is necessary to compensate the moving background. In this paper a real-time technology of compensating moving background based on matching weighting function and multi-resolution analysis that can be used in high-capability FPGA is presented. The technology uses affine parameters model to acquire the global motion parameter and block matching critera to acquire global motion displacement estimation of sparse sample points set. A matching weighting function is presented to eliminate the influence of moving objects. In order to meet the requirement of real-time processing and accuracy the multi-resolution method and parallel pipeline structure are used in the system.In this paper based on the object block character, the difference between background moving and object moving, and the short-time stationarity of image sequences, a multilevel detection technology of video moving objects in moving background is presented using Markov random field (MRF). Firstly it uses the difference image of adjacent frames, which the background moving has been compensated, to acquire the possible existing region of moving objects based on Neyman-Pearson principle. Then a region growth algorithm...