A target detection method based on point cloud completion

Abstract

The application relates to a target detection method based on point cloud completion, and belongs to the technical field of laser radars, and solves the problem of low target detection accuracy in the prior art by using laser radar point clouds. The target detection method comprises the following steps: acquiring a laser radar point cloud as an effective point cloud, and acquiring a current frame camera image and a previous frame camera image which are synchronized with the laser radar point cloud; according to the current frame camera image and the previous frame camera image, the effective point cloud is expanded to obtain an expanded point cloud; the expanded point cloud is pretreated to obtain a plurality of to-be-classified point cloud clusters; each to-be-classified point cloud cluster is input into a point cloud feature extraction and classification network which is pre-trained, so that the class prediction probability of each point in each to-be-classified point cloud cluster is obtained; and the target class of each to-be-classified point cloud cluster is determined according to the class prediction probability of each point in the to-be-classified point cloud cluster. Higher-accuracy identification of a target is realized.

Images