Calibration method and system applying camera and scene association of intelligent driving

Abstract

The invention discloses a calibration method and system applying a camera and scene association of intelligent driving. The method comprises steps of monitoring a single or multiple target objects in the camera view field in advance through a camera which is arranged on an automobile windshield and has an adjusted view angle; when each target object walks in the camera view field, carrying out imaging through the camera, and using the position and height information, detected by a detector, of each target object in an image coordinate as calibration data; and according to the calibration data, through offline calculation or online learning, updating a calibration model used for representing the position and height relations of the target objects. According to the invention, by calibrating the target objects appearing in the camera view field through the camera installed on the automobile, structure distribution properness of pictures and targets in the imaging pictures are mainly calibrated; and the calibration method is advantageous in that the calibration method is simple, rapid, accurate and free of environment limitations.

Description

technical field [0001] The present invention relates to the technical field of camera calibration, in particular to a camera and scene correlation calibration method and system for applying intelligent driving. Background technique [0002] In the prior art, road scene perception technology based on visual sensors is widely used in automatic driving, semi-automatic driving and advanced driver assistance systems (Advanced Driver Assistance Systems, ADAS) (collectively referred to as intelligent driving). The main function of the camera is to continuously perceive the visible light image information around the car, especially in front, and send it to the processor for scene analysis and understanding. After relevant decision-making processing, it will be fed back to the user or directly involved in the driving system. [0003] A single on-board camera realizes the mapping from the 3D real world to the 2D image, that is, Px=u(1); [0004] Among them, P is a projection matrix s...

AI Technical Summary

Problems solved by technology

[0012] 1. The calibration and calculation process is cumbersome and requires more human intervention;

[0013] 2. Only intermediate results such as internal and external parameters of the camera can be obtained, and additional calculations are required to obtain the final target-image mapping relationship;

[0014] 3. It is difficult to realize the combination with the automatic detection algorithm, and the detection algorithm cannot be accelerated;

[0015] 4. Unable to detect changes in camera installation configuration

Images