Vehicle three-dimensional information detection method, device, equipment and readable storage medium

Abstract

The invention provides a vehicle three-dimensional information detection method, device, equipment and readable storage medium. The method includes: using the actual scene image collected by a monocular camera as the input of a pre-trained deep learning model, and obtaining the output of the deep learning model information, the output information includes four binary classification values ​​indicating whether the four vehicle side edges are visible and the image coordinates of the intersection points of the four vehicle side edges and the road surface; the vehicle orientation is determined according to the four binary classification values; the determined vehicle orientation is composite When facing, the vehicle heading angle is calculated according to the image coordinates of the intersection point of the visible vehicle side edge and the road surface, the actual length and width of the vehicle, and the internal and external parameter matrices of the monocular camera. Compared with labeling the vehicle heading angle, it reduces the difficulty of labeling the training images, and detects the three-dimensional information of the vehicle based on the single frame image of the monocular camera, thereby reducing the cost and improving the detection stability.

Description

technical field [0001] The present invention relates to the technical field of automatic driving, and more specifically, to a vehicle three-dimensional information detection method, device, equipment and readable storage medium. Background technique [0002] Understanding the surrounding traffic environment is critical for autonomous vehicles. When an autonomous vehicle is driving, it not only needs to detect surrounding vehicles, but also needs to detect three-dimensional information such as the position and orientation of the vehicle. At present, for the 3D information detection of vehicles, binocular cameras or depth cameras are used for detection, but the cost is high; there is also the use of labeled data from lidar to train the vehicle 3D information detection model, and then use the training in the actual scene. The model detects the three-dimensional information of the surrounding vehicles based on the scanning data of the lidar, but the lidar has the problem of low...

AI Technical Summary

Problems solved by technology

At present, for the 3D information detection of vehicles, binocular cameras or depth cameras are used for detection, but the cost is high; there is also the use of labeled data from lidar to train the vehicle 3D information detection model, and then use the training in the actual scene. The model detects the three-dimensional information of the surrounding vehicles based on the scanning data of the lidar, but the lidar has the problem of low stability

Images