Automatic two-camera calibration method and system

Abstract

The invention provides an automatic two-camera calibration method and system. The method comprises the steps of controlling PTZ cameras for shooting a monitored region according to the preset movement step length in the horizontal direction and the movement step length in the perpendicular direction to obtain PTZ images, and recording shooting parameters needed when each PTZ image is shot, wherein the overlapping region of the previous and later PTZ images shot by the PTZ cameras is larger than the preset overlapping range, and the shooting parameters of the PTZ cameras include the current shooting multiples of the PTZ cameras, movement values of the PTZ cameras in the horizontal direction and movement values of the PTZ cameras in the perpendicular direction; extracting feature points of each PTZ image and feature points of a wide-angle image; matching the feature points of each PTZ image with the feature points of the wide-angle image to obtain a matched pair of each PTZ image and the wide-angle image; according to pixel coordinates of the feature points of the PTZ images and the wide-angle image in the matched pairs, obtaining a mapping relation between the coordinates of the pixel points in the wide-angle image and the shooting parameters of the PTZ cameras. According to the automatic two-camera calibration method and system, the calibration process is simple, and the calibration result is high in accuracy.

Description

technical field [0001] The invention relates to the technical field of camera calibration, in particular to an automatic calibration method for dual cameras and an automatic calibration system for dual cameras. Background technique [0002] At present, in the field of security monitoring, the need to obtain clear surveillance video images as the basis for surveillance and law enforcement is constantly increasing. Traditional single-camera surveillance systems cannot directly obtain clear images of moving targets in the distance in large scenes, but manually pass through the target objects. Zooming in for close-up magnification, this monitoring method will cause a huge waste of manpower, and often leads to the failure of monitoring personnel to report due to fatigue. Therefore, the application engineering cost of the existing dual-camera monitoring system is relatively high, and it is only used in special and important places, and cannot be widely used in the current urban se...

AI Technical Summary

Problems solved by technology

[0005] One is manual calibration. Select a pixel in the wide-angle image as the target, write down the image coordinates, and then control the dome camera so that the target point of the gun camera is located in the center of the dome camera screen, and then calibrate each pixel in the wide-angle image. Marking, forming a point mapping table of one-to-one correspondence between the gun camera pixel of the monitoring system and the motion position of the dome camera, but the process of actual application is relatively cumbersome, especially in large scenes, the process of one-by-one marking is very huge

[0006] One is the geometric mapping model method, which requires a sufficiently accurate camera imaging model and environmental geometric model, and adopts multi-reference point zoom conversion and fixed-focus parameter model parameters. The accuracy of the calibration mapping depends on the accuracy of the model parameters. The parameter solution The process is relatively complicated, and it is easy to cause the solution parameters to not converge, and the accuracy of the calibration cannot be guaranteed

[0007] The other is the data fitting method, which collects the information between the wide-angle pixel position information and the camera motion parameters, and uses polynomial approximation or least square method to obtain the mapping relationship. In the traditional technology, more than three pairs of wide-angle bolt images are used. The pixel coordinates and the horizontal and vertical heights of the dome camera are paired with the least squares method to solve the mapping transformation matrix. The internal parameters of the camera are obtained through the distortion correction method, and more than three pairs of the paired points are matched by using the sift / surf feature of the image. algorithm, but the biggest problem with this method is that there will be a large deviation in the large outdoor scene and the scene with a single texture and a large number of repetitive textures. The accuracy of the parameter solution parameters will affect the accuracy of the calibration. If the There are many scenes similar to squares and repetitive textures in the monitoring area of ​​each security and traffic area in the whole city, which is difficult to implement

Another is to use a large amount of offline data for the calibration process, which requires a large amount of data collection for fitting, and the amount of calculation is relatively large.

Images