As the transmission line point cloud in complex environment is different due to the structure and point cloud density and structure of each power element, the noise point distribution around the power element is irregular and uneven, so the traditional denoising method is difficult to denoise uniformly. To address this problem, this paper proposes a coarse classification-based denoising algorithm for transmission line 3D point cloud data. Firstly, a location-based point cloud extraction method is introduced to coarsely classify the tower point clouds, and the coarsely classified power line point clouds and feature point clouds are obtained based on the Euclidean clustering segmentation algorithm (euclidean cluster extraction). The coarse-classified tower point clouds and feature point clouds are denoised using the density-based spatial clustering algorithm (DBSCAN) and statistical filtering algorithm with adaptive thresholding, respectively. For power line point clouds, this paper proposes a coordinate transformation algorithm based on the principal component analysis (PLE-PCA) algorithm of power line direction estimation, which can realize the point cloud denoising under the sparse and uneven distribution of power line point clouds by effectively amplifying the distance between noise points and power line point clouds. The study uses UAV dense matching point cloud data and UAV laser point cloud for denoising experiments. The comparison experimental results with radius filtering algorithm and statistical filtering algorithm show that the method has good robustness in the complex environment of transmission line laser point cloud and dense matching point cloud with dense noise, mixed power elements and noise points, low density and uneven distribution of power line point cloud.