Generating Joint Cameraand LiDAR Features Using Cross-View Spatial Feature Mapping for 3D Object Detection

摘要

A method and apparatus for generating fusion information based on multi-sensor data for 360-degree surrounding object detection and recognition are presented. The multi-sensor data-based fusion information generation method for 360-degree surrounding object detection and recognition tasks proposed in the present invention includes the steps of obtaining a feature map from a multi-sensor signal using a deep neural network, and integrating the obtained feature map. It includes the steps of converting to a dimensional coordinate system and generating a fusion feature map for performing recognition using the converted integrated three-dimensional coordinate system.