Target vehicle position and orientation detection method and system and storage medium

Abstract

The invention provides a target vehicle position and orientation detection method. The method comprises the steps: step S10, collecting a front view image of the vehicle through a vehicle-mounted camera; step S11, preprocessing the front view image collected by the vehicle-mounted camera; step S12, performing image motion compensation on the front view image according to vehicle-mounted inertial measurement equipment; step S13, according to an inverse perspective transformation rule, converting the position of each target vehicle in the front view after image motion compensation into a top view; and step S14, inputting the top view into a pre-trained convolutional neural network to obtain position and orientation information of each target vehicle. The invention further provides a corresponding system and a storage medium. According to the invention, the distance and orientation detection precision of the target vehicle based on vision can be greatly improved.

Description

technical field [0001] The present invention relates to the technical field of intelligent driving, in particular to a method, system and storage medium for detecting the position and orientation of a target vehicle. Background technique [0002] In the intelligent driving of the car, it is necessary to detect the distance between the front and rear targets according to the driving environment. At present, the main method of vision-based target detection is: in the front view, obtain the two-dimensional rectangular frame (bounding box) of the vehicle target in the image according to the CNN convolutional neural network (YOLO, SSD or Faster-rcnn, etc.). Its general method flow is as follows figure 1 As shown, the steps include: firstly carry out preprocessing operations such as resize on the input front view image; then perform neural network reasoning on the preprocessed front view to obtain possible two-dimensional rectangular boxes (bounding boxes) of all target vehicles;...

AI Technical Summary

Problems solved by technology

[0004] First, the distance calculation of the vehicle target position is not accurate, and the error is large

In the front view, the lower boundary of the two-dimensional rectangular frame of the vehicle target is often not the position of the ground point of the vehicle, which causes a large error in the detected target vehicle position distance relative to the real value, and the distance between the target vehicle and the vehicle The farther away, the greater the error of the measured distance value

[0005] Second, the posture orientation of the target vehicle cannot be effectively detected

In the front view, it is often only possible to detect the two-dimensional dimensions of the width and height of the vehicle target, and it is difficult to detect the orientation of the target vehicle.

[0006] Therefore, the existing vehicle target detection based on the front view has the disadvantages that it is not easy to measure its motion posture, and the position and distance error is relatively large.

Images