Researches On Obstacle Detection On Transmission Lines Based On View Synthesis

The impact of high-voltage transmission lines damage arising from the long mechanical tension in the field, electrical flashover and materials aging, which has a serious thread of the operation of power system; it must be regular inspection of transmission lines. The main method of transmission line inspection includes manual inspection and helicopter inspection at present, the manual inspection methods which has intensive labor, low accuracy and a high risk for staff because of the special environment of transmission line; the helicopter inspection methods which has increase the efficiency and accuracy of inspection, but increase the technical difficulty and higher operating costs. The transmission line inspection robot is currently a hot research which is to replace the manual and helicopter inspection method, the navigation technology is one of the key technologies of transmission line inspection robot, the obstacle recognition for inspection robot based on multi-view reconstruction is depth studied in this paper, the specific content of the following:1) After analysis and comparison of the traditional transmission line inspection robot obstacle recognition methods, an obstacle recognition method based on multi-view reconstruction is proposed in this paper, first the obstacle recognition process is divided into two parts include online process and offline process, the multi-view reconstruction is done and the multi-angle virtual profile image library is build for the common obstacle on the transmission lines in offline parts; the obstacle image is taken by any angle online, then extract the obstacle contour, at last the contour is matched with the virtual image library which build offline to achieve the obstacle recognition. The complex calculations if completed offline and simple calculation online for this obstacle recognition method, which can recognize the obstacle quickly and eliminate the impact of shooting angle.2) According to the process of the proposed obstacle recognition method, the multi-angle obstacle image reconstruct and obstacle recognition is the key problem to solve. In the process of obstacle image reconstruction, an improved combination of adaptive gradient magnitude and morphological filter algorithm, an improved adaptive multi-window gradient magnitude edge detection algorithm and an improved stereo matching algorithm based on parallax interpolation is proposed in this paper based on gradient magnitude. At last, the multi-view reconstruction algorithm for obstacle contour and the obstacle recognition method based on the point statistical is proposed.The standard image is used in the simulation experiment for the algorithm proposed in this paper and the real obstacle images is used to test the efficient of obstacle recognition, the preliminary results show that the good recognition results is got by using the method in this paper, through the study of several algorithms proposed in this paper, the complete identification of obstacle recognition method is formed. Key words: Obstacle recognition Multi-View synthesis Gradient magnitude...