High Accuracy Detection of Image Features in Vision Model Calibration

In computer stereovision, the extraction of 3D space information and shape reconstruction from 2D stereo images depend largely on the vision model. Calibration of the vision model parameters has been and continues to be, one of the most heavily investigated topics in computer stereovision. Aiming at plane calibration sample with grids, a high accuracy method for detecting image features is presented. By using sub-pixel edge detection and edge tracking etc, small square edges and cross edges of the calibration sample image are distinguished, then center lines of the cross and focuses of the small squares are determined, and accurate image features are detected. The method presented is applied to images with simulated noise and real images. The experiment results are given at the end of this paper.