Research On Power Tower Detection Using Fast Efficient Heuristic Spectral Clustering Algorithm

An effective way to solve the problem is the flying robot. A flying robot is able to acquire a large quantity of images of overhead power line by carrying visual sensor system. Therefore, only through inspecting the key objective-power tower, can the large quantity of inspection information be tackled faster. Furthermore, it can also offer reliable location information of the power tower. Given those mentioned above, the research proposes rapid and efficient power tower inspection algorithms under different inspecting backgrounds.The paper analyses the characteristics of the environment of power tower and view angle where images are captured, and carries on power tower inspection methods toward the two typical environments. For power tower inspection under simple background, we proposes a rapid and efficient heuristic clustering algorithm. By combining advanced heuristic hill-climbing algorithm and spectral clustering algorithm together, it not only overcomes the disadvantage-heavy dependency on initial clustering center-in classical spectral clustering algorithm, but also reduces the amount of computation and improves computing efficiency. Tlnougli applying the rapid and efficient heuristic clustering algorithm to experiments on power tower images under simple background, the experimental result proves that it is steady, fast and effective, posing effective performance for power tower images in simple background.However, texture in complicated background is rich. In such circumstance, clustering algorithm cannot segment area of inspection effectively. In order to solve inspection problem of power tower under complicated background and inspect power tower in images precisely, the paper adopts global self-similarity descriptor as characteristic mode of power tower and effectively represents the repeated texture features in power tower images. To improve clustering speed of descriptor construction, this, combined with extraction process of rapid and efficient heuristic clustering algorithm’s descriptor, makes power tower inspection more effective and precise. The paper takes different real images of power tower as experimental objectives and eventually validates the accuracy, efficiency and low space complexity.Inspection method in the paper lays a solid foundation for automatic process and data analysis. By using automatic power tower inspection, both power tower and its attached facilities can be effectively extracted from a large number of inspection data for its preservation and repairing. The intensity of labor is reduced, and the quality analysis and processing speed of inspection images are increased. Thus, it is promising to promote the automatic overhead power line inspection research and application.