Moving Object Shadows Detection Algorithm Research And Implementation

Inter-frame difference algorithm is a simple,classic moving object detection algorithms,very practical,and has applied in many occasions.Gaussian mixture model(GMM)is the most practical method and the most commonly used in moving object detection,this technology is mainly composed of three parts,including background model,initialization,and renewal,which are also the most key parts of GMM.On the basis of the predecessors' analysis,this paper analyzes and summarizes the development and application of background model in moving object detection,which is based on the dual background model in complex environment.This paper discussed the current popular from the theory and implementation of moving object detection algorithm: inter-frame difference algorithm and GMM,and the effect of foreground object detection of these two methods in the different standard test sets is analyzed.This paper summarizes these two method do not completely solve the problem of moving object detection in a complicated environment.At the same time,the traditional Gaussian mixture model can do nothing,when the initial input frame image has a moving object.Based on this,firstly,this paper introduces background extraction and moving object are obtained by using improved Gaussian mixture background models,and then the previous moving object and moving object of frame difference are effectively integrated into a moving foreground object we are interested in,finally,we extract the background region under the moving foreground object.It can greatly restrain the influence of the complex environment mentioned above.In addition,due to the moving object covering the source light,it forms the shadow,which makes the extraction of the foreground object will contain the shadow,which affects the detection effect of moving object.In this paper,a simple and effective method is proposed that utilizes the texture information difference between the shadow and the gray image background,and based on the color space to detect the difference between the foreground object and the shadow and the correlation between the background and the shadow,to extract the foreground object shadow region,and eliminate the vision of the shadow area,and then detect the true moving object.