Research And Application Of Camera Calibration In Vision Measurement System

In vision-based measurement, the relative position change of the camera and the outside world results in the frequent change of the internal and external camera parameters, eventually causing the measurement errors. RAC two-step method and plane calibration method can realize online camera calibration, but owning to noises and space constraints in metallurgical industry, these methods can't be used in the industrial scene for its accurate calibration equipments demand.In this paper, aiming at the above-mentioned problems, a camera calibration method based on cylindrical measurement bar ring is presented. Trough the recognition of the feature point coordinates and the establishment of a reliable calibration model can realize the parameters calibration of camera and online-update of the measurement bar position relation. This method can be used in the tundish molten steel level measurement and the error is less than 5mm. The main research contents of this paper are as follows:1. Research of camera calibration based on the three-dimensional measuring bar targetBased on the three-dimensional measuring bar camera calibration target model, the paper has proceeded theoretical analysis and mathematical deduction and realizes the calibration function of internal and external parameters and then analyzes the number of data resource, accuracy, algorithm, constraint of parameters which effect the calibration results. At the same time, the coordinate conversion relation establishment of the point on the measurement bar realizes the transformation from the image coordinate to the real coordinate.2. Corner reliability recognition and feature point extraction based on three-dimensional measuring barIn the process of the production of the chess-board calibration target, the corner transfer region will show the phenomenon of coincidence and separation, resulting in the failure of the corner location. To solve this problem, this paper compares the commonly used corner detection methods and selects SV operator, to ensure reliable corner only to be identified, which achieves the extraction of calibration data and the positioning error is less than 1.2 pixels. 3. The applications of the camera calibration method in the tundish molten steel level measurementIn the tundish molten steel level measurement, the three-dimensional measuring bar camera calibration method is adopted. Establish coordinate conversion relation and reverse coordinate projection the points of the measurement bar can realize the level measurement. At the same time, according to the shift and rotation in the actual work, the paper corrects the pose in real time through the calibration of the external parameters and ensures the reliability of measurement.In a word, the three-dimensional measuring bar camera calibration method can overcome the deficiency of the noises and space constraint in the metallurgical industry. The process of the calibration is simple and reliable and the requirements of the environment are low, thus the calibration method has a good applicability and application prospect in the parameters measurement in harsh metallurgical industry environment.