The invention discloses a power corridor scene classification method based on LiDAR point cloud features, and the method comprises the steps: firstly, determining a target power corridor scene, and collecting the laser scanning data of the scene; secondly, preprocessing such as denoising the collected original LiDAR point cloud data; determining the shape and size of a neighborhood according to the point cloud space topological structure of different ground object targets, defining reasonable features, carrying out feature extraction, and constructing a single-point feature vector; and finally, classifying the point cloud of the target scene by using the labeled point cloud data through a random forest classifier to obtain the point cloud of the tower, the power line, the vegetation, the building and the ground. According to the method, the defects of low precision and poor data adaptability of a traditional scene classification method are overcome, and automatic classification of power corridor scenes is realized.