The Research And Implementation Of Computer Vision Camera Calibration Systembased On MATLAB

Camera calibration is one of the important research visual inspection, photogrammetry, computer vision and other areas, it is through a series of three-dimensional geometry of space on the physical characteristics of the corresponding point in the two-dimensional image of the relationship, calculate the final results with optimized camera internal and external parameters. In the image measurement or computer vision, precise calibration parameters are very critical step to determine its calibration accuracy and stability of the results algorithm to produce camera work directly affects the accuracy of the results.In this paper, MATLAB programming language to develop a camera calibration system, software system function modules include: calibration template production, image processing, camera calibration, image distortion correction and the corrected image matching stitching.The image processing part in removing the noise and edge-preserving win-win bilateral filtering, the largest division within the variance between the adaptive improvement determine the conditions of subpixel edge gray matrix coordinate extraction and least squares fitting algorithms to extract the center oval Asia pixel coordinates.For camera calibration is required to ensure that the calibration was smooth and rigorous precision calibration object manufacturing and maintenance not only difficult and costly problem, we use the LCD as a calibration template to solve the usual practice in many engineering applications, or the need for a lower cost The high precision flat calibration object problems. Now that LCD monitor production process is relatively mature, pure flatness deviation of less than 0.05 um, pitch accuracy up to 0.01 mm, much higher than ordinary printed cardboard precision, so in the world coordinate measurement accuracy can meet the high calibration and measurement accuracy. By automatically obtain the computer screen resolution to draw any size of custom precision circular feature points as the calibration and automatically get the physical characteristics of the dot center coordinate space, the successful solution of such problems. Camera calibration algorithm on the basis of Zhang Zhengyou joined tangential distortion and improve parameter optimization algorithm to optimize the calculation, and further improve the calibration accuracy.After the use of the calibration parameters on the distortion resulting image taken under the same conditions can be corrected for image matching and stitching. In this paper, image rotation, scaling, brightness changes remain stable under the SIFT(Scale Invariant Feature Transform) feature matching algorithm early match, and then increase the number of iterations using RANSAC(Random Sample Consensus) algorithm for matching points go wrong, Finally, the issue fade out the right way given a weighted average fusion algorithm eliminates stitching good marks.