Modeling and Identification Method of 10kV Overhead Lines Applicable to Autonomous Inspection

There is a vast territory and a comprehensive coverage of overhead distribution lines in China, characterized by many points and a wide range of distribution. Vehicle-mounted PTZ inspection is a future development trend with broad application prospects. This paper proposes a power distribution line recognition method based on the improved DeepLabV3+ network for inspection vehicles. Given the problem of spatial information loss in the original model, an improved semantic segmentation method based on the DeepLab model compensates for the detailed information, reduces the probability of missing due to the small pixels occupied by the power lines in the image, and improves the semantic segmentation accuracy. For the phenomena of crossing and bending in the power distribution lines captured during the intelligent inspection process, a method for overall identification of power distribution lines is proposed, which improves the robustness of recognition and can determine whether there are phenomena such as occlusion and missing inspection through the continuity of identified lines. Finally, this paper uses examples to simulate and verify the rationality and effectiveness of this method in power distribution overhead line target recognition.