Power transmission line point cloud semantic segmentation method and device based on multi-source feature extraction

The invention relates to the technical field of high-voltage power transmission line three-dimensional reconstruction, in particular to a power transmission line point cloud semantic segmentation method and device based on multi-source feature extraction. According to the method, the thought of multi-source data features is utilized, and the method comprises three links of down-sampling feature extraction, up-sampling interpolation and full-link segmentation. Wherein the down-sampling feature extraction link is mainly used for sampling and grouping point clouds to obtain central point and domain point data; then respectively extracting high-dimensional features of the center point, the domain point and the relative position information; and finally, splicing the extracted feature information. The up-sampling interpolation link is mainly used for carrying out interpolation and feature cascading on the data. The full-link segmentation link mainly carries out point-by-point classification according to feature in