A Novel Technology Research Of Detecting And Tracking Moving Targets

Detection and tracking of moving objects is an important project in the field of computer vision, which is applied widely in intelligent video monitoring, human-computer interaction based on video, vision-guided robot, virtual reality, automatic transmission, etc.This thesis has researched detecting and tracking technique of moving objects.On the basis of analyzing the principle and property of the Gabor filter, optical flow computation method based on Gabor orientation gradient is proposed. First, the image is filtered by Gabor filters in different orientations, and the edge images are extracted in different orientations. Secondly, the contour image is reconstructed from the edge images. As a result, the grey distribution of the complex background can be eliminated except its edges. Thirdly, gradient optical flow arithmetic is used to compute optical flow fields, and the orientation of optical flow computation is the same as that of Gabor filter. The experiments results show that the optical flow noise of the grey distribution in background can be reduced, the orientation of optical flow is accurately captured, and the moving objects can be easily identified from the background in optical flow fields.In the respect of object tracking, the algorithms of particle filter (PF), kernel particle filter (KPF) and particle swarm optimization (PSO) are researched. The adaptability between PSO and KPF is discussed. The tracking algorithm of optimized kernel particle filter by particle swarm (PSO-KPF) is proposed. First, the optimal solution of particles tracking in previous time is used to the speed equation of particles in current time. Then, the perturbation equation of KPF is optimized according to the principle of PSO. As a result, the particles, far from true state, are driven to the high likelihood area. The simulation experiments and the video tracking experiments show that the performance of PSO-KPF is superior to that of PF and KPF, and the precision and speed in tracking are improved.