基于激光扫描的鱼眼相机三维标定方法

发布时间：2018-12-29 07:21

【摘要】：由于鱼眼相机成像存在较大的畸变,采用二维标定板的方法难以在图像边缘区域获得准确可靠的角点,从而导致标定精度下降,而传统的三维标定法存在标定场建造复杂,特征点数目有限等问题。为此,提出一种基于激光扫描的鱼眼相机三维标定方法。该方法首先通过激光扫描仪获取室内标定空间的三维点云图,然后利用尺寸不变特征变换匹配方法得到点云图与待标定相机照片的对应点的匹配关系,并进行分块随机抽样一致性(RANSAC)筛选,再根据对应点的图像坐标和物理坐标进行三维RANSAC筛选,估算最终的鱼眼相机内外参数。与经典的张正友标定法及其改进方法相比,该方法能够获得更多的有效特征点,使标定精度明显改善,从而较好地实现鱼眼相机的畸变校正。该方法简便、精确,有广泛的适用性。
[Abstract]:Due to the large distortion in the imaging of the fish-eye camera, it is difficult to obtain accurate and reliable corner points in the edge region of the image by using the two-dimensional calibration plate method, which leads to the decrease of the calibration accuracy, while the construction of the calibration field is complicated in the traditional three-dimensional calibration method. Finite number of feature points and so on. Therefore, a three-dimensional calibration method for fish-eye camera based on laser scanning is proposed. Firstly, the 3D point cloud image of the indoor calibration space is obtained by laser scanner, and then the matching relation between the point cloud image and the corresponding point of the camera photograph is obtained by the size invariant feature transformation matching method. The consistent (RANSAC) selection of block random sampling is carried out, and then the 3D RANSAC screening is carried out according to the image coordinates and physical coordinates of the corresponding points to estimate the final internal and external parameters of the fish-eye camera. Compared with the classical Zhangzhengyou calibration method and its improved method, the proposed method can obtain more effective feature points, improve the calibration accuracy and achieve the distortion correction of the fish-eye camera. The method is simple, accurate and widely applicable.
【作者单位】：上海大学机电工程与自动化学院;中国科学院宁波材料技术与工程研究所;
【基金】：国家自然科学基金(61273276) 国家科技支撑计划(2015BAF14B00) 浙江省自然科学基金(R13F020014)
【分类号】：TP391.41

【相似文献】