作者: Zhengyou Zhang
DOI:
关键词: Camera matrix 、 Camera auto-calibration 、 Distortion (optics) 、 Artificial intelligence 、 Digital camera 、 Homography (computer vision) 、 Computer vision 、 Mathematics 、 Pinhole camera model 、 Camera resectioning 、 Feature (computer vision)
摘要: A digital camera is calibrated by establishing the coordinates of at least four feature points a pattern mounted on planar surface. At two, and preferably three or more, images are captured different, non-parallel orientations using camera. The image pattern's then identified in images. closed form solution can be employed to derive all intrinsic extrinsic parameters needed provide calibration. Essentially, known corresponding used compute homography for each image. Then, process that estimates analyzing homographies associated with Finally, computed from homographies. However, effected various noise sources which could affect accuracy process. If higher called for, maximum likelihood inference either more accurate first estimate, refine derived solution. radial distortion caused lens also concern, further refined taking into account this distortion.