THREE-DIMENSIONAL OBJECT DETECTION FRAMEWORK BASED ON MULTI-SOURCE DATA KNOWLEDGE TRANSFER

摘要

A three-dimensional object detection framework based on multi-source data knowledge transfer. Image features extracted by an image feature extraction unit are output, so that a target of interest selection unit outputs point cloud data of a target of interest to a point cloud feature extraction unit according to the image features; the point cloud feature extraction unit extracts point cloud features from the point cloud data; then, in a knowledge transfer unit, the image features learn the point cloud features, and parameters of the image feature extraction unit are updated; a three-dimensional target parameter prediction unit updates parameters of the image feature extraction unit and the point cloud feature extraction unit according to the image features and the point cloud features; and finally, the updated image feature extraction unit re-extracts the image features to the three-dimensional target parameter prediction unit, and the three-dimensional target parameter prediction unit calculates and outputs the three-dimensional parameters according to the image features, and therefore, the detection precision of three-dimensional object detection based on a two-dimensional image is improved.