A Point Cloud Superimposed Sampling Method Based on Farthest Point Sampling Strategy

As self-driving technology continues to become more popular and used, the object detection task in the perception system of self-driving vehicles has gained more and more attention, in which the LiDAR sensor can collect point cloud data. Data sampling in point cloud information processing is a fundamental and crucial task in object detection. This paper combines the network framework of deep learning and proposes a point cloud superimposed sampling (PSS) method based on the farthest point sampling method. Compared with related research work on a public point cloud benchmark, good point cloud classification results are obtained, proving the effectiveness of the proposed sampling method.