The Electric Power System Transmission Line Detection And Tracking Based On Visual

With the rapid development of Chinese electric power industry and the need of theconstruction of the intelligent power, the inspection of the transmission line based on visual isbecoming a hot spot of research. In this paper, we study of transmission lines’ detection andtracking based on visual and its purpose: one is the foundation to identify the damage andfracture of transmission lines; the second is to explore the UAV to track transmission line andcomplete automatic navigation based on visual. Algorithms in this paper are implemented andverified by using MATLAB2009a development tools under the win7operating system. Themain contents are as follows:（1）Firstly, this paper study and implement three kinds of edge detection algorithm:the multi-scale linear enhancement filter, Unit-linking PCNN filter and the mean ratio of linedetection operator. Considering difficulty to matching algorithm parameters, the speed ofalgorithm and the actual effect of transmission line detection, this paper determined using themean ratio of line detection operator in the follow-up work（2）Secondly, this article studies line primitive extraction algorithm of the Houghtransform and phase grouping and analyzes their advantages and disadvantages. The gradientpositioning of phase grouping is not very accurate to make the support linear area oftransmission line fragmented. Aimed at that point, an improvement method is proposed in thispaper. And this paper realizes the combination of phase grouping and Hough transform. Thealgorithm has the advantage of phase group and Hough transform, extracting the low contrastedge and having the stability of Hough transform. It also improves the speed of algorithm.（3）Next, this paper studies the post-processing of lines primitive. We first need to lineprimitive fitted to a straight line, and we analysis the two methods of least squares andRANSAC. RANSAC algorithm can eliminate noise interference and gain the ability to robustlinear models. Then on the basis of line, k-means clustering method can exclude theinterference of linear background. For the fault of transmission line connection, this paperadopts the method based on evidence theory to connect the breaking transmission line. Finally it is able to present a more complete transmission line.（4）Finally, this paper studies the tracking methods of the discrete Kalman filter andextended Kalman filter, and attempts to establish a moving target model of transmission line.We establish the interesting tracking region using the Kalman filter and realize the trackingfor video image sequences. We also extent the hypothesis for extended Kalman filter stateequation, but due to the limited experimental conditions, we do not realize the experimentalsimulation for this method.