Calibration Algorithm Optimization In Structured Light Surface Measurement Technology

Science optical measurement technology is a non-contact method, with high precision and high speed, and it has been widely used in industrial inspection, heritage digitization, medical science, CG, agriculture, etc. But its accuracy still can not be comparable to coordinate measuring instrument. The key factor restricting the precision of three-dimensional structured light surface scanning device is camera calibration precision and the variation of camera parameters due to environmental factors. Since the accuracy of target feature sub-pixel coordinates have important implications on the camera calibration results, this paper first analyzes the common target feature sub-pixel coordinates extraction algorithm, and compares two most commonly used circular feature point calibration target sub-pixel center coordinates extraction algorithm: gravity method and ellipse fitting method, the experimental results show that they have the same effect. The main method to improve the calibration accuracy is optimizing calibration results with bundle adjustment, but bundle adjustment can not be directly applied to optimize common surface scanning system cameras. The main way to eliminate environmental factors which affect the measurement accuracy of the three-dimensional structured light surface scanning system is frequent recalibration, which is lack of effective online detection method and cumbersome. Therefore, an improved bundle adjustment algorithm to optimize camera calibration result and an online detection method with self-correction algorithm to improve the accuracy of three-dimensional structured light surface scanning device precision is of great significance.Sparse bundle adjustment(SBA) has been widely used in photogrammetry and machine vision. However, the direct use of SBA to optimize the result of Zhang's single camera calibration result will get different groups of camera internal parameters and distortion parameters(camera parameters). Therefore, the mathematical model of SBA is constrained to produce unified camera parameters, which can lead to changes in the structure of the Jacobian matrix. Then, this paper presents a constrained sparse bundle adjustment(CSBA) algorithm with a new block matrix partition strategy to improve the efficiency of solving sparse linear equations, which can be used to optimize the camera calibration precision of Zhang's method. And simulation experiments are used to verify that CSBA has better tolerance to pixel coordinates error than SBA. Besides, practical experiment shows that CSBA can be used to optimize the camera parameters and positionparameters and get higher reconstruction accuracy than existing SBA optimize algorithm.There is no practical way to monitor the accuracy of three-dimensional structured light surface scanning device and the calibration process is complex. In this paper,corresponding points are matched based on multi-frequency heterodyne unwrapping of two perpendicular directions, and the accuracy of the camera parameters are monitored through the distance from corresponding points to the epipolar line, so the accuracy can be monitored online. Subsequently epipolar line constraint and levenberg-marquardt algorithm are combined to optimize camera parameters. In order to improve the phase matching accuracy, a rectangular area phase similarity match method is proposed to remove isolated noise and an adaptive threshold adjust algorithm is used to fulfill phase fine matching. In order to match the corresponding point as accurately as possible, a new method based on probability distribution principle for removing mismatching points is proposed. At last the effectiveness of the algorithm is experimentally verified, and the camera parameters' changes in the operational process can be monitored and thus certain corrections can be made to optimize the camera parameters, then the measurement accuracy can be maintained within a certain scope.