Point cloud objective recognition method combining SHOT features and ESF features

During the process of obtaining a point cloud, various problems, such as noise, occlusion, and incompleteness, will affect the recognition accuracy of the object. This paper proposes a point cloud 3D object recognition method combining SHOT features and ESF features to identify the objects in complex point cloud scenes accurately. The model is recognized based on the template matching method. According to the corresponding group and Hough voting method, we can determine the matching key points and the global features are calculated based on the rotation invariance characteristic of point clouds. The experiments show that the proposed method is, on average, 15% more accurate than traditional feature descriptor based on identification methods, and our approach also presents better robustness to noise.