Defect recognition for unmanned aerial vehicle inspection based on global feature fusion of LiDAR images

In order to ensure the safe and stable operation of transmission lines and accurately obtain the defect data required for their operation and maintenance, a defect recognition method for unmanned aerial vehicle inspection based on laser radar image feature fusion was proposed. Using laser image global feature point detection, and then using random histogram method to reduce the feature point vector set, global feature fusion was performed on the bottom layer, middle layer, and high layer respectively. Based on the laser image of the transmission line after global feature fusion, clustering algorithm was used to achieve intelligent recognition of transmission line drone inspection defects. According to the time and other parameters contained in the laser image captured by the drone during inspection, the defects of the transmission line were located to obtain the actual location of the transmission line defects. The experiment shows that this method can effectively fuse the global features of laser images, with a maximum global feature fusion degree of 0.99. The accuracy of defect recognition in different environments remains high, and the localization accuracy is close to 100%. In comparison, the average processing time of this method is also the shortest, only 5.6 seconds, which verifies its superior performance and application effect in the field of defect identification and localization in transmission line inspection.