Automatic high-precision recognition method of planar coding grid for Camera Calibration

Abstract:
In order to improve accuracy and efficiency of feature point extraction for camera calibration, a planar target is developed by combining a planar grid with three red positioning circle marks and displayed on the LCD. A new method is proposed to extract the nodes by utilizing the characteristics of cross neighbor area and combining with the sub-pixel extraction method. In the same time, a coding method is developed by constructing coding coordinate system and bidirectional stack structure to match the nodes on the images and object points. The proposed method is then applied to calibrate a nonmeasurement web-camera. The experimental results show that the detection accuracy of feature points can be improved and accurate camera calibration can be acquired with the back-projection root mean square error (RMS) of the checkpoints in the images within 0.13 pixel. The method can also be extended to other automatic applications which adopt planar grid as the reference target.