Object Detection And Tracking For Active Camera

Computer vision is an important field in computer science, and the object detection and tracking in image sequences is a very important research topic in the field of computer vision. The traditional object detection and tracking is studied under the fixed camera, along with the diversification of monitoring, appearing more and more the moving camera, such as vehicle equipment, robot, handheld devices, PTZ camera. When the background motion, object detection and tracking method in fixed background is no longer valid. So the research for object detection and tracking in mobile scenarios is very necessary. This paper aims to detection and tracking research for mobile camera, discusses the principle of object detection and tracking and effective algorithm, has important academic significance and application value.Study on the present status of object detection and tracking is reviewed. For the mobile camera movement under the background due to the difficulty of object detection, we used one kind of method based on the background characteristics block matching to compensate the movement of background, so as to realize the detection of objects using frame difference method; and the object value of multiple object detection based on the gradient characteristics, the value of the binary gradient feature BING a simple calculation, can well meet the requirements of real-time systems.For the mobile camera under the object tracking, the TLD algorithm is improved based on the popular, the original TLD algorithm for object recognition rate is not high, large deformation or drift problem, improved learning algorithm, to improve the recognition rate by increasing the tracking module to achieve accelerated algorithm of learning; and through the boxfilter matrix and the optimal template matching search mode, the algorithm can improve the high complexity of computation deficiencies. Finally, the experiments verify the feasibility of the system, through the analysis of experimental results, achieved good results.