Research On Moving Object Detection And Shadow Suppression Algorithm In Video Sequence

The intelligent video surveillance system not only distinguishes interesting target objects from complex video frames,but also analyses useful information from the video,which improve the intelligent perform of the traditional video surveillance system.In the field of the intelligent video surveillance,the moving object detection is the basis of video analysis,and it plays a key role in the subsequent research of motion recognition,tracking and behavior understanding.At the same time,the object shadow has the same motion properties as the object and is detected together when the object are detected,which reduces the accuracy of the detection result.Therefore,the shadow elimination of the moving object becomes the most important task.This article takes the moving object in the video and analyzes the detection method of the moving object and the shadow elimination algorithm.Then we improve these method to find the better effect of the moving object detection in the video.The paper mainly obtained the research results in the following aspects.1.This paper research the principle of classical object detection algorithm which includes optical flow method,frame difference method and background subtraction method,and we make corresponding experimental tests.Three kinds of classic object detection algorithms are compared from the aspects of algorithm complexity,acquisition of image information,applicable conditions and existing problems by experimental analysis and discussion of advantages and disadvantages2.This paper focuses on the research of the Gaussian mixture model algorithm research and we propose an improved algorithm which is based on Gaussian mixture model.Aiming at the disadvantage of high complexity in the mixed Gaussian model,an algorithm which combines the frame difference method and the Gaussian mixture model is proposed.Firstly,the three-frame difference method is used to detect the foreground,and the ratio of the foreground and the image is obtained.Then the dynamic condition in the image is judged by the threshold.If there is the moving object in the image,the mixture is processed using a Gaussian model,otherwise the background is updated.Aiming at the disadvantage of the fixed background update rate,the background update rate is adjusted adaptively through different rate.Experimental results show that accuracy of improved algorithm is 4% higher than traditional algorithm and the proposed algorithm significantly improves the program running time of the traditional Gaussian background model algorithm.The proposed algorithm overcomes the problems caused by the fixed background update rate of the Gaussian mixture model.3.In order to suppress the shadow of the moving object better,the commonly shadow elimination algorithm based on HSV space is improved.First,the foreground is extracted by the Gaussian mixture models,and then combined with the shadows detected by the HSV space to obtain a moving object region.because of the disadvantage that the foreground misdetection is caused by the instability of the brightness ratio threshold,the texture feature is combined with the OTSU to extract part of the moving object.The part of the extracted moving object is finally combined with the moving object region obtained through the HSV space so as to suppress the shadow of the moving object in the video.The experimental results show that this method preserves the complete moving target and eliminates the shadow effectively.Compared with other commonly used shadow elimination algorithms in quantitative analysis,the proposed algorithm has much better increment of about 8%.And it can satisfy the real-time needs.In this paper,the research results of moving object detection and shadow elimination can achieve accurate extraction of moving object in video,which has great significance for subsequent video analysis and processing.