Traffic vehicle monocular positioning method, device and equipment and storage medium

Abstract

The invention relates to the technical field of driving assistance, and discloses a traffic vehicle monocular positioning method and device, computer equipment and a storage medium, that is, after a monitoring image acquired by a monocular camera is acquired, a target vehicle in the monitoring image can be found through a target detection algorithm; the method comprises the following steps: acquiring a monitoring image of a target vehicle, obtaining vehicle size feature information of the target vehicle in the monitoring image, inputting the vehicle size feature information as a to-be-detected sample into a distance classification model which is based on a deep learning algorithm and has completed training, obtaining a distance classification result, and finally, taking the distance classification result as a depth estimation result. Therefore, depth estimation can be realized only by combining with a monocular camera, the hardware cost is greatly reduced, an SLAM (Simultaneous Localization and Mapping) method does not need to be adopted, the requirement on computing power can be reduced, a positioning result can be quickly obtained, and the method is particularly suitable for being applied to a vehicle-mounted computer. In addition, based on the improvement of the Yolov3-tiny network, the problem of inaccurate small target detection can be solved.

Description

technical field [0001] The invention belongs to the technical field of driving assistance, and in particular relates to a traffic vehicle monocular positioning method, device, equipment and storage medium. Background technique [0002] Today, 3D reconstruction technology is a hot research field in the direction of computer vision. From a 2D plane image to a 3D three-dimensional space, it is not only an ordinary scale transformation, but also involves depth estimation, which is relatively difficult at present, because the camera imaging process is a process of converting 3D space coordinates into image 2D coordinates. , and if the 3D coordinates are reversed from the 2D coordinates, there will be a problem, that is, there is a scale factor S in the process from the image coordinate system to the camera coordinate system that cannot be determined, because the 2D image does not involve the concept of depth. [0003] Nowadays, people have higher and higher requirements for inte...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problems of high hardware cost and high computing power requirements in the existing traffic vehicle positioning technology, the purpose of the present invention is to provide a traffic vehicle monocular positioning method, device, computer equipment and computer-readable storage medium, which can only need Combining with a monocular camera can achieve depth estimation, greatly reducing hardware costs, and does not need to use SLAM methods, which can reduce computing power requirements and help to quickly obtain positioning results. It is especially suitable for application in on-board computers and is convenient for practical application and promotion.

Images