The combination of vision and unmanned aerial vehicle is an economic and effective solution for transmission line inspection. The key in the process of transmission line inspection is real-time extraction and recognition of transmission lines regardless of the complicated background. The horizontal and vertical disturbances are numerous in the complicated natural background. At the same time, transmission line crossing and different shooting angles of unmanned aerial vehicle lead to the phenomenon of overlapping lines, which makes it more difficult to extract and track transmission lines. Therefore, a three-step synthetic extraction algorithm for transmission lines is proposed in the article. Firstly, in view of the edge extraction problem of the image, a complementary edge feature extraction algorithm combining the Sobel operator with the Log operator is proposed to solve the problems such as discontinuous and uneven edges and low-detection efficiency caused by the general Canny operator for edge detection. Secondly, in view of the line feature extraction, a coarse-fine directional optimization algorithm is proposed. The horizontal and vertical interferences of the extraction of transmission lines under complex background and the problem of line crossing in the image are eliminated. The algorithm has higher detection effect and efficiency than Hough transform. Finally, a similarity determination mechanism is proposed to merge the adjacent lines, which avoids the loss of the tracking line caused by overlapping lines in the image. A quadrotor unmanned aerial vehicle is used to carry out the experiment of transmission line detection in the article. The experimental results demonstrate that the three-step method of line extraction is more effective and efficient than the current algorithms.
Feature Extraction of Gear Fault Signal Based on Sobel Operator and WHT
