Research On Algorithms For Moving Object Detection In Video

In computer vision and intelligent video surveillance and other fields, the higher levels, such as moving object classification, tracking and behavior understanding, depend heavily on the results of moving object detection in video. Moving object detection is one of the fundamental tasks in video analysis. This paper focuses on algorithms for moving object detection in video. A new algorithm for detecting moving object based on adaptive background subtraction and symmetrical differencing has been brought in this paper. The main contributions are as follows:1.Considering poor SNR of real image sequence, An adaptive weighted mean filter approach is proposed. The filter fully uses the local correlation of neighboring pixels with respecting to the center pixel. Then the image will be enhanced with histogram amendment.2. Two different types of dynamic threshold approachs are proposed in the paper, they are used in moving object detection in video.①A fast adaptive threshold approach for image is proposed in the paper, according to making object and background separated completely. This approach can obtain a good impact of image segmentation.②Another dynamic threshold approach is proposed in the background subtraction in moving object detection. The idea of this approach is to change the threshold dynamically according to the changes of light when obtaining two images. This approach can adapt to the complex changes of the environment scene and overcome the impact of light changes effectively.3.Several classical background modeling algorithms are widely researched and analyzed based on extensive experiments, and the approachs are compared each other in speed and accuracy. Finally, the comparison conclusions are presented.4.A novel algorithm for moving detection, which employs adaptive background subtraction and symmetrical differencing method, is proposed in this paper. A modified selective updating model is proposed as the reliable adaptive statistical background updating method, and we combine the background subtraction with symmetrical differencing to detect moving information. After the motion detection operation, morphologic filtering and connected region area measurement are introduced to suppress the noise and solve the background disturbance problem, the area filling algorithm is used to fill the small hole in the detected object. Finally the moving objects are extracted reliably. A conclusion of the whole text and an expectation of future work are given at the final part of this paper.