Three-dimensional human ear recognition method based on 3DHarris key points and optimized SHOT features

The invention discloses a three-dimensional human ear recognition method based on 3DHarris key points and optimized SHOT features. The method comprises the following steps: (1) carrying out target detection on a two-dimensional side face image with background information by utilizing a detector based on a Faster R-CNN framework, obtaining a square human ear region and extracting a three-dimensional point cloud of a corresponding region; (2) filtering the human ears by using voxel grids; (3) registering the human ear to be measured and the human ear in the standard posture by using an iterativeclosest point algorithm, enabling the human ear to be measured to have a standard vertical upward posture, converting the human ear to be measured, placing the converted human ear in a standard coordinate system, and removing redundant areas such as skin and hair by using cuboid template cutting to obtain a normalized human ear; (4) performing key point detection in the three-dimensional human ear point cloud according t