Design And Implementation Of Intelligent Deviation Correction System For Power Inspection Drone

With the development of science and technology,UAVs are playing an increasingly important role in the inspection of high-voltage lines.At present,although many companies integrate real-time dynamic carrier phase difference technology into the autonomous inspection of UAVs,when UAVs land after high-altitude inspections and inspections,there will still be target objects that deviate from the center of the screen and landing points that deviate from the specified The situation of the location.In this paper,the main goal is to improve the accuracy of UAV target object tracking and autonomous landing.By designing a UAV autonomous deviation correction inspection system based on target detection,a high-performance fully autonomous UAV high-voltage line inspection is realized.The main work of this paper is as follows:First of all,the visual aided landing technology is studied.This paper optimizes the conventional image preprocessing,and uses the adaptive threshold to perform the binarization operation on the image to filter out the noise in the binarized picture.For the Sobel edge detection operator,the gradient matrix of eight directions is introduced.Improve the Hough transform circle detection to make it suitable for the situation where the circle radius of the UAV is constantly changing during the landing process.Compared with the pre-processing algorithm before optimization in the landing mark detection,the X-axis landing deviation is reduced by 53%,and the Y-axis deviation is reduced by 32%.In the research of target detection framework,in order to meet the hardware conditions of UAV embedded devices,lightweight improvement is made on the basis of YOLOX target detection framework,Backbone is replaced by Mobile Net V3,and Coordinate attention is introduced in the output feature layer.Ghost Module is introduced,and Depthwise convolution is used instead of ordinary convolution in Neck to reduce the number of parameters and computation of the model.Alpha-DIOU loss function is introduced to improve the accuracy of bounding box regression.Compared with the original YOLOX＿tiny,the new model increases the m AP0.5:0.95 index by 2%,but the amount of parameter calculation is greatly reduced,and the running speed on Nvidia NX can reach56 FPS.The experimental results show that the system can effectively solve the problem of deviation correction of patrol UAV.