Moving Object Detection In Dynamic Background

In the area of computer vision research,motion object detection is one of the research directions.Its purpose is to detect and extract the moving objects in the video sequence from the surrounding environmental information.The results of the detections can be used in object identification,target tracking,behavior analysis and many other areas of research.The application areas involved from the military video surveillance to civilian security monitoring.Because the background is non-static,such as the environmental disturbances of trees,rain,snow,waves and so on,changes in lighting and camera movement,the object detection will be caused by greater interference.Eliminating the negative dynamic background Influence has always been the key focus and difficulty of research.Started with the causes of the dynamic background,this paper made two aspects of the study:1.For the dynamic changes caused by the environment,the research is based on the background elimination method in this paper.We propose a background model for kernel density estimation based on pixel gray value and optical flow value,which is on the basis of the background modeling method and the optimization of the initial results.Based on the good adaptability of kernel density estimation and optical flow method in dynamic background,the initial results of the detection are analyzed.Based on the characteristics of the image processing time with hidden Markov random field,an observation field model based on kernel density estimation is proposed to optimize the initial results.Through the simulation experiment,the test results verify the effectiveness of the proposed method.2.For the dynamic background generated by the camera movement,in order to solve the problem that when the object is relatively static,using the background elimination method will lead to the gradually disappearance of the relatively static goal,this paper proposed a method of motion estimation of the objects.Firstly,the region of interest is extracted from the initial test results.Then,the motion estimation of the interested multiple regions is carried out by using the simple advantages of CT algorithm in single target tracking.Finally,a region adaptive method is proposed.It is estimated that the object's detail information in the target area is proposed by using the K-means algorithm to complete the detection.The experimentalresults verify the effective solution to the problem of disappearance of the object.