Data set establishment method, vehicle and storage medium

Abstract

The present invention relates to the field of automatic driving technology, in particular to a method for establishing a data set and a vehicle. The method includes acquiring millimeter-wave radar images and lidar images; performing space and time calibration on millimeter-wave radar and lidar; The deep neural network of the lidar image and use the deep neural network to generate the target estimation result of the lidar; project the target estimation result onto the matching millimeter-wave radar image as a pseudo-label of the millimeter-wave radar image; The radar target confidence is generated from the radar local signal; the millimeter-wave radar dataset is established according to the target confidence and pseudo-labeling. Compared with traditional manual labeling, the present invention can realize automatic labeling with the help of laser radar and improve labeling efficiency, and based on the deep neural network of model integration, it can realize target recognition with high recall rate, filter out false positive target detection frames, and improve labeling efficiency. quality.

Description

technical field [0001] The invention relates to the technical field of automatic driving, in particular to a method for establishing a data set, a vehicle and a storage medium. Background technique [0002] Millimeter-wave radar has become an integral part of the Advanced Driver Assistance Systems (ADAS) of automobiles. The radar can also work normally and has strong robustness. The work of traditional millimeter-wave radar is to use digital signal processing algorithm to extract the position and velocity information of the target, and establish a data set based on a series of radar points or point clouds. However, the generated radar point or point cloud does not have information such as the size and attitude of the target. In automatic driving, it is usually necessary to use the pose information including the position, depth, size and angle of the target to complete the work of environment perception. However, since the millimeter-wave radar datasets applied to automatic...

AI Technical Summary

Problems solved by technology

However, the generated radar point or point cloud does not have information such as the size and attitude of the target.

In automatic driving, it is usually necessary to use the pose information including the position, depth, size and angle of the target to complete the work of environment perception. However, since the millimeter-wave radar datasets applied to automatic driving are usually annotated manually, the annotation efficiency is extremely high. Low

[0003] Although other sensors, such as cameras, can also be used to mark mmWave radar, the accuracy of camera estimation for pose information is low

Images