Research On Moving Object Detection Based On FMP In Moving Camera

Moving target detection is a hot topic in computer vision.And with the development of the society,many scholars focused on moving target detection under moving cameras,rather than whether stationary cameras.With human vision,this thesis focuses on the change between successive frames to detect moving target.In order to detect moving objects under moving cameras,we proposed a moving object detection algorithm based on FMP(fail matching point).First,we extracted the stable points of failure matching based on the matching of stable spatial feature points.With the target moving,the target and the background handover area change,which leads to the stable point matching failure matching.Then,the moving region of the moving target is determined according to the stable points of the failure matching.Finally,we realized the detection of moving targets by combining the local image segmentation algorithm.The main parts of this thesis are as follows:(1)We extracted FMP based on SIFT feature by comparing different spatial feature point extraction algorithms.(2)This thesis gave the definition and calculation method of FMP.(3)The FMP distribution is analyzed and the three frame processing method is proposed to obtain stable FMP.Based on the analysis of FMP distribution,we extracted stable FMP by three frame processing method.(4)We gained the moving target region based on matching vectors of the successful matching points in the neighborhood of FMP.(5)With the FMP of moving target region and grabcut(image segmentation),this thesis detected moving object.(6)We studied the detection of multiple moving objects based on FMP in the target area and successful match points.Experiments showed that accurate FMP can be extracted under different movement modes of the camera,and this algorithm not only extracted single targets,but also extracted multiple targets,whether or not the targets were occlusion.And the algorithm can also distinguish different moving targets.