| The detection and tracking technology of moving target involves a number of areas consisting of video processing, image processing, pattern recognition, intelligent algorithm and cognitive systems. Researching on the cable-laying area under surveillance, by the means of image processing, the prevailing ships can be detected and tracked automatically. It helps a lot on reducing the probability of accidents and ensuring the submarine power grid operate steadily.In this paper, the basic techniques of image preprocessing and common method of target detection have been studied. According to the ship’s features which are slow motion and a fixed direction, the Gaussian mixture model has been applied to set up the background, in order to overcome the defects that background subtraction is sensitive to external environment, especially the light. So the method of adaptive background subtraction has been applied to detect the target.In this paper, canny operator has been applied to extract the edge of the positive or negative samples from ship video and seven-stage Hu invariant moments are calculated. This feature has fully reflected the edge information. By analyzing the extracted characteristics, two classifiers have been designed to recognize ships based on support vector machine and relevance vector machine respectively. What’s more, classifier parameters have been optimized so as to improve the accuracy. On the basis of classification evaluation and identification results, RVM classifier has been applied to recognize the target.A method combining the edge orientation direction histogram(EOH) and CamShift algorithm has been applied in this paper. CamShift algorithm tends to lost its target when the target color is in the same with the background. The method proposed in this paper can overcome the shortcomings and it do well in ship’s tracking despite of the complex conditions.So the method proposed in this paper brings an accurate, effective, and well real-time tracking. |
PDF Full Text Request