Target detection method based on millimeter wave radar prior positioning and visual feature fusion

Abstract

The invention discloses a target detection method based on millimeter wave radar prior positioning and visual feature fusion. The method includes: acquiring millimeter wave radar point cloud data and camera images simultaneously based on the calibrated millimeter wave radar and a vehicle-mounted camera; The wave radar point cloud data is subjected to spatial three-dimensional coordinate transformation to project to the camera plane; based on the projected millimeter wave radar point cloud data, multiple anchor samples are generated according to the preset anchor strategy, and the speed and distance weights of each candidate area are obtained. The final anchor sample; by fusing the RGB information of the camera image and the scattering cross-section intensity information of the millimeter-wave radar point cloud data, the characteristics of the final sample are obtained; the characteristics of the final anchor sample are input into the detection network to generate the target in the scene Category and location information. The method of the present invention fully utilizes the characteristics of the return information of the millimeter wave radar to optimize the existing detection framework, and improves the accuracy of on-road target detection in an all-weather scene.

Description

technical field [0001] The invention relates to the technical field of automatic driving, in particular to a target detection method based on millimeter-wave radar prior positioning and visual feature fusion. Background technique [0002] Target detection in the road scene is the key technical link to realize the vehicle's automatic driving perception, so as to ensure that the vehicle driving perception system determines the environment around the vehicle, avoids collisions with other vehicles, pedestrians and other targets, and predicts the execution of the auxiliary decision-making system in advance to realize the vehicle Safe autonomous driving on the road. [0003] The automatic driving perception system recognizes the surrounding targets, which can be used to identify the targets in the current scene by extracting the features of the visual image acquired by the camera, which is very similar to the human perception through the visual system. It's easier for people, but...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to overcome the deficiencies of the prior art, and propose a target detection method that fully utilizes the multi-mode fusion of millimeter-wave radar information and vision. The main feature of this method is to fuse millimeter-wave radar point clouds and camera images Target detection, using millimeter-wave point cloud as a supplement to image space information and enhancement of semantic features, using images to make up for the defect of low sampling density of point cloud, using fusion information to enhance visual information Targets in complex road scenes such as rain, fog, and night Robustness of Detection Algorithms

Images