Detecting And Tracking Moving Objects In UAV Video Based On Optical Flow Analysis

Moving target detection and object tracking, is one of the hot and difficult topics in computer vision, which has broad application prospects in intelligent transportation, security monitoring and other military fields. Unmanned Aerial vehicles (UAVs) which has advantages of high maneuverability, high resolution, convenient hiding and flexible operation, is mainly used in military surveying and mapping fields ,day and night aerial reconnaissance, battlefield surveillance, battle damage assessment and etc. Therefore, UAVs which is equipped with video sensors to track and analyze the moving ground target that has important practical significance and theoretical value in military area. However, because of the characteristics that the moving video sensor is moving with the high-speed UAVs simultaneously, the complexity of the background in the video image sequence and the diversity of moving target information, which makes the problem of treatment target detection and object tracking becomes even more difficult. The research purpose of this paper is to bring optical flow technology into the moving target detection and tracking field. The major work and achieved results now has been completed are as follows:1. Describe the basic principles of optical flow and common calculation methods, and upon this, analyze and summarize the application plan that how the optical flow technology is used in motion estimation and object detection and tracking, which laid a theoretical foundation for the following chapters of the application of optical flow technology.2. In order to eliminate the displacement of the image background which is caused by camera movement , this paper uses motion estimation and global motion compensation,which is based on the harris feature points matching, to proposes a method which uses optical flow technology to calculate background motion estimation, build background motion parameter model, and finally compensate the global motion; Comparing experiments confirms that the optical flow method have great advantages of rapid calculating and great practicability.3. Moving target detection method is briefly analyzed and difference moving target detection method is realized after background motion compensation. The content of this chapter emphasizes on the level method that is used in contour segmentation and combines the optical flow and level set to facilitate moving target detection and segmentation, which achieves satisfying results in both still and moving background conditions .4. Background motion estimation, global motion compensation and the detection and segmentation of moving target is realized through the optical flow method in dynamic background. And based on the method, this chapter uses continuous adaptive mean shift(Camshift) and Kalman filter prediction method to track target and thereby, achieves continuous tracking of both single and mutiple targets.