Research On Moving Target Detection Algorithm Based On Background Modeling

As an important branch in the field of image processing,mobile target detection is the basis of intelligent analysis techniques such as behavior recognition and target tracking in video systems such as intelligent video surveillance systems.In many computer vision-related science and technology is a very important basic technology.The moving target detection technology analyzes and predicts the moving trajectory of the moving target by acquiring the moving parameters(acceleration,position,etc.)of the moving target,and then processes the obtained data.In order to achieve different processing of moving targets(including detection,location,identification,classification and tracking,etc.).As we all know,optical flow method is a kind of important algorithm in moving target detection technology,but the current optical flow method is easy to be affected by the change of light and background movement,and the detection of moving target under dynamic and complex background is not very accurate.In order to solve the above problems,this paper aims at the traditional Lucas-Ka The nade optical flow method is optimized and improved by combining the inter-frame differential method,meanshift clustering algorithm and morphological processing,so that the improved Lucas-Kanade Optical flow algorithm can be used in simple environment.Or complex environment under the detection results have made great progress.At the same time,the improved algorithm also shortens the execution time to a certain extent,and has the ability to resist the noise interference such as the change of light.The main work of this paper is as follows:1.The main drawback of Lucas-Kanade Optical flow algorithm is that it has a large amount of computation and high time complexity.In order to improve this defect effectively,this paper uses inter-frame subtraction to reduce the range of images to be processed,thus reducing the computational complexity of the algorithm.Speed up,less time consumption: this article compares five inter-frame difference method,finally chooses the inter-frame difference method with the least time complexity and the fastest detection speed.The detection data is processed in the early stage,and the noise of the smaller area in the detection image is removed by removing the small figure in the MATLAB,and then the original image is cutaccording to the processed binary image.The contours of moving objects in two successive frames are obtained,which greatly reduces the computational complexity.2.The results of this algorithm are further optimized by using the function,median filtering,canny operator of deleting small targets,which reduces the background noise points and improves the accuracy of detection.1)removing the background noise by deleting the small objective function for morphological processing.2)smoothing the detection region through median filter operation,3)calculating the contour boundary of the region containing "foreground" by canny operator.3.Aiming at the interference of background movement to Moving target detection,the background points are deleted by mean shift clustering algorithm,and the feature point sequences of input image are filtered and filtered effectively.In order to reduce the computational cost and running time of mean shift algorithm,the: mean shift clustering algorithm divides the filtered feature point vectors into a similar class,calculates all kinds of center points,and then selects the center point of the class which contains the most number of points and regards it as the background point.Remove similar points from the vector diagram.4.A variety of improved algorithms based on Background modeling are implemented in this paper.These improved algorithms are compared with the improved algorithms in terms of running time and detection accuracy.The experimental results show that the improved algorithm achieves better moving target detection and location under the premise of ensuring the running speed.