Missing pin detection method for overhead transmission lines based on SPD and ASFF

In transmission lines, bolts and pins play an important role in fixing wires and connecting fittings. During the inspection of transmission lines, pins on transmission towers are important detection objects. In this paper, a missing pin identification method based on SPD and ASFF is proposed. Firstly, in order to increase the prominent defect features, the annotation information of normal category bolts is added to the database. Secondly, in order to extract richer feature information, the SPD module is used to replace the original simple convolution module. Finally, the ASFF structure is applied to reduce gradient backpropagation. According to the experimental results, the improved model’s mAP of missing pin detection increased by 7 percentage points. The proposed method achieves rapid identification of missing bolts in overhead transmission lines, greatly improving the detection accuracy.

I. INTRODUCTION

The industries and the livelihoods of people depend heavily on the safe and reliable operation of overhead transmission lines. The most common inspection techniques nowadays include manual inspection, inspection from a helicopter, inspection from a robot, inspection from a drone, etc. [1]. Compared with other inspection methods, drone inspection has the advantages of wide detection coverage, Due to the advantages of low operating requirements and high cost performance, drone inspections combined with manual inspections are currently widely used in actual power system work to improve work efficiency and safety. A large number of visible light images will be generated during the drone inspection process. By inspecting the visible light images, the bolts and pins in the transmission lines are missing. However, manual inspection has problems of low efficiency and high labor cost.