Moving Target Detection Based On Improved Frame Difference Method And Background Subtraction Method

Moving object detection is one of the most important and difficult points in the field of computer vision and image processing. It is the basis of moving object tracking and recognition. Frame difference method, background subtraction method and optical flow method are commonly used to detect moving targets. This thesis for the frame difference method and the background difference method to launch the research, the two methods were improvement and fusion, and the fusion algorithm is applied to moving target detection in complex scenes, so as to improve the moving object detection accuracy in complex scenes.Firstly, when the moving object and the moving background are similar, the three frame difference method can not be used to obtain the complete contour of the moving object. This thesis in two frames difference method based on using Canny edge detection algorithm to extract moving object contour information, and then with the three frame difference method combined with, the improved three frame difference method to improve the integrity of the contour of the moving target under complex scene is obtained.Secondly, because the traditional Gauss mixture model is the random number in the sample points, the detection result is not ideal. Based on Gaussian mixture model background difference method is improved, using the K-means clustering algorithm to fast obtain clustering center, then the the point as a Gaussian mixture model background update initialize mean parameters, thereby improving in complex scenes with moving target detection accuracy.Finally, in order to improve a variety of scenarios of the adaptability of the moving target detection. In this thesis, the design of the a moving object detection algorithm, the algorithm combines the improved frame difference method and the improved background subtraction method overcome the limitations of separate detection algorithm based on complex scenes on the environment and improve the of complex scene moving target detection adaptability.