Research On Anti-interference Algorithm Of Motion Detection

As people's requirement of safety increasing, video monitoring is developing towards digital, network and intelligence direction, which may have wide uses in daily life, business and national defense. The application of motion detection technique can greatly improve the performance of intelligent surveillance system by automatic target detection. Motion detection technique can judge the existent of abnormal phenomenon and decide whether to trigger alarm equipments. Once abnormal conditions occur, the system will give an alarm immediately to the operator and start to make a recording of the scene, which can save manpower and storage space.In actual application, many kinds of interference(light,weather.) may reduce the sensitivity and accuracy of motion detection because of complex environment. False alert often occurs, and brings people a lot of trouble. When the rate of false alert reachs a fairly high level, video monitoring system will lose its meaning.The article is devoted to studying moving targets detection, shadow elimination, illumination processing in static scene.This paper proposed a new algorithm entitled "Double rectangles" after researching some common moving object detection algorithms. In the proposed algorithm, the moving region is first detected according to the two difference images of three adjacent images. Then it extracts the target with Background Subtraction algorithm in the detected moving region which is a rectangle and the target has precisely located in. By using the gray characters of the vidio image, a good background can be extracted from the former N frames.Then use the Surendra algorithm to update the background according to the current frames in real time. The Double rectangles algorithm combing the Background Subtraction algorithm and the Frame Difference algorithm cleverly may extract the moving object totally and also reduce the influence of noise at the same time.Bacuse Otsu algorithm only adapts to the bimodal histogram of images, this paper proposes a new image threshold segmentation algorithm combing Otsu algorithm and unimodal histogram algorithm to segregate target from background in the process of image segmentation. In the algorithm, a threshold is first determined primarily by Otsu algorithm which is used to judge whether the histogram of difference image is unimodal or bimodal. If it is unimodal, we need recalculate the threshold by unimodal histogram algorithm; otherwise this one does. Experiments show that the threshold is more precise and the segmentation result is better.Based on HSV color space, this paper implements shadow elimination, taking advantage of characters which shadow influences luminance component rather than chrominance component and combing with Double rectangles algorithm proposed by this paper.This paper integrates moving detection algorithm and shadow elimination algorithm and make them nested with each other. It kills these two birds with one stone. It will either simplify the procedure or eliminate the shadow effectively.In the end, this paper brings forword the corresponding improvement algorithms to realize the light change cancellation according to different light change condition. The result of experiments show that the motion detection algorithm proposed by this paper has good robustness of environment change. It can eliminate shadow fairly good and reduce the influence of light change to the detection result of moving objects. So, it has a high application value.