Defect diagnosis technology based on multi-spectral point cloud

To increase the ability to detect defects, 2D image RGB and temperature information was fused with 3D point cloud by calibrating cameras and LiDAR. According to the segmented area in 2D images, point cloud was semantically divided into different groups. Visual appearance defects and abnormal temperature distribution in the segmented point cloud can be recognized with less misdiagnosis. The point clouds collected at different times were compared to realize the abnormal appearance of equipment components. This application showed an improvement of the reliability of power equipment and reduction of the work pressure of operation and maintenance personnel.