The Study Of Moving Detection Technology Based On 2-D Entropy And MRF

In this paper, a study on the problem of motion objects analysis based on images sequence is presented. As a key technology especially of the objects segmentation and objects tracking in the field of machine vision, motion objects estimation has many potential important applications in a wide range of fields including military affairs, process of industrial control, medicinal research, traffic control, aero navigation, etc. The research is based on moving region segmentation and tracking, and its purpose is to detect moving objects in dynamic image sequence, track trajectories of objects, therefore the motion parameter such as displacement, speed and acceleration can be obtained. Then moving objects detection is a important basic work for all motion objects analysis.The main content of this paper is as follows:Firstly, the purpose of the research of the motion objects analysis in the domain of computer vision is introduced. The research of development and the correlative regions are discussed. The content of this research is presented.Secondly, the method of 2 -D (dimension) entropy is discussed and the application of 2-D entropy in image processing is also presented. The method of 2-D entropy is used for image restoration, edge detection which leads to a satisfactory result. More over, a new method based on statistical feature for segmenting moving objects from monocular image sequence regardless of complex, changing background is discussed. Based on this, every image frame is mapped from spatial domain to statistical domain. Then, a max 2-D entropy framework is proposed for segmenting moving objects in statistical domain.Thirdly, a method of Markov random field (MRF) is also presented in image processing, and its relative models solving is discussed. A novel method based on MRF and boundary MRF for segmenting moving objects from monocular image sequence proposed. An algorithm of mathematical viewpoint statistical regularization based on MRF is presented, and is proceed by mapping the algorithm onto the energy model. The model based on spatial-temporal neighborhood system is given and the cost function is fixed by the model. ICM (interact condition method) method is adopted to get the better result.Finally, introduce the definition of the moving edge .A method of moving edge that detected based on image edge is analyzed. So we propose a new method to detect moving objects. It is a method based on MRF to segment object and background .We can get the moving edge from the segment result.