This paper addresses the automatic extraction of street lamps from 3D point clouds using unmanned aerial vehicle (UAV) data, proposing a sample matching layered extraction method. Firstly, the original point cloud is divided into three layers: lamp head layer, lamp pole layer and ground layer by cloth simulation filtering(CSF) algorithm, and hierarchical clustering is carried out. Then, the point cloud of lamp pole is extracted by sample matching algorithm. Finally, based on the relative position constraint between the lamp head point cloud and the center of the lamp pole, the lamp head point cloud is extracted, thus completing the extraction of the whole street lamp. Our experimental results show that in different road environments, the correctness, completeness and F1 indices of street lamps extracted by the proposed method are not less than 90%, which can achieve efficient and complete extraction of the street lamp.