提出了一种以优化点云数据中平面的平面度和平面面积为目标的内参数标定方法, 针对不同情况选用区域生长算法或改进随机采样一致性(RANSAC)算法进行平面提取, 并通过提出的单纯形和人工蜂群混合优化算法(NMS-ABC)进行参数解算, 实现了DIY系统的内参数标定.利用内参数已知的仿真数据实验验证了NMS-ABC混合算法用于内参数快速求解的有效性, 利用不同场景获取的多组实测点云数据进行了标定实验, 结果表明: 标定后的平面内点百分比和整体平面度均有所提高, 并且当平面数不少于3时标定结果更稳定可靠; 进行了点位精度测试, 经内参数标定后DIY扫描仪的点位测量精度得到提高且在不同水平角度分辨率条件下优于3 mm@4 m; 人为增大安置误差, 通过对比标定前后的点云及平面提取结果, 进一步直观验证了本文所提标定方法的正确性和有效性.

A calibration method of intrinsic parameters was proposed, which aimed at optimizing both the flatness and the area of planar patches detected from the point cloud. Either region growing algorithm or improved Random Sample Consensus (RANSAC) algorithm was chosen to extract planes according to different cases. A hybrid optimization algorithm (i.e. NMS-ABC) integrating the Nelder-Mead Simplex algorithm with the Artificial Bees Colony algorithm was proposed to calculate parameters, by means of which the intrinsic calibration of the DIY system was successfully implemented. Simulation experiments were performed using synthetic data based on known intrinsic parameters, which demonstrated the efficiency of NMS-ABC hybrid algorithm. Actual calibration experiments were conducted with multiple groups of point cloud data acquired in different scenes. Results indicate that both the percentage of planar inliers and the total flatness are improved after calibration. Besides, the calibration results will be more stable and reliable when the number of detected planes is not less than 3. Position precision test was conducted, and position precision of the DIY scanner was improved to the level of 3 mm@4 m after calibration. By increasing installation errors artificially and then contrasting point coulds as well as extracted planes before and after calibration, it is visually validated that the proposed calibration method is correct and effective.