Monocular computer vision image calibration method and its application

Big error of locating control points of anamorphic vision image calibration cause it difficult to calibrate of dynamic image automatically and accurately. Accordingly, a new calibrating method based on pattern matching is put forward. It locates the control points accurately; and builds the polynomial modeling of the anamorphic image, chooses the optimum polynomial order and calibrates the image distortion with bilinear interpolation. Apply the calibrating method to steel level measurement in continuous casting tundish base on compute vision; the method achieves quick and accurate calibration of image in steel level measurement. The results show that the error between measuring and real steel level is about 20 mm before calibration; then it comes to 0.6 mm after calibration. It provides an important link of image preprocessing to monocular computer vision measuring system.