| Regularly inspecting high voltage power transmission line and repairing dilapidation are essential to keep power supply reliable. At present, overhead power line mainly is manually inspected by workers, which is dangerous, with high working intensity, and costs much. Therefore, It is necessary to develop robot for power line inspection, which aims at the improvement of work efficient, the decrease in cost and the guarantee of the operation line. This thesis is based on this background to do an in-depth study on target detection and tracking method for overhead power lines inspection robot.This paper summarizes the most commonly used target detection and tracking method in the video, and pointed out the advantages and disadvantages of various methods and application. Then, on the basis of this study, made the following improvements:For target detection, taking into consideration the complexity of the actual situation, this paper proposed a method which combined the background subtraction and the Adaboost machine learning algorithms. In this method, the video image sequence to do background subtraction operation firstly, then use the Otsu method to binarization images and the use of morphological filtering to extract moving regions, and finally used the classifier which was trained by the Adaboost algorithm to detect the region, which get the video target shock hammer.For target tracking, using the method of based on detection and color features tracking. First, take a brief introduction of the color model, color histogram and color probability distribution, and gives experimental results. Then, the focus of the classic Meanshift algorithm, proved its deficiencies. Finally, the proposed algorithm based on improved Camshift target tracking. The algorithm is initialized using the results of target detection in the search window, thus avoiding the use of color characteristics to describe the objectives of unity, to improve the tracking accuracy. |
PDF Full Text Request