Multiconstraint Spatial and Temporal Calibration of Rotating Line Structured Light Vision Sensor
Multiconstraint Spatial and Temporal Calibration of Rotating Line Structured Light Vision Sensor
复制标题
旋转线结构光视觉传感器的多约束时空标定
DOI：
10.1109/tim.2021.3107612
复制
发表时间：
2021
影响因子：
5.6
通讯作者：
Xiong Rong
中科院分区：
文献类型：
--
作者：
Han Fuzhang;Zhang Qunkang;Fu Bo;Yang Tong;Wang Yue;Xiong Rong
给定运动的线结构光视觉传感器(LSLVS)因其结构简单、扫描速度快、功耗低而被广泛应用于许多工业领域。其中一个应用是旋转LSLVS作为三维激光雷达，它适用于功耗有限的任务，如行星探测任务。由于LSLVS的运动特性和时间同步问题，使得LSLVS与转台之间的精确标定成为此类系统的关键。在这项工作中，我们的目标是对旋转的LSLVS进行两个阶段的空间和时间校准。在第一阶段，根据不同方位采集的图像数据计算外参数，在此过程中，LSLVS保持不变，以避免时间偏差造成的误差。此外，为了具有较高的校准精度和鲁棒性，在该阶段加入了多个约束条件。然后，在第二阶段，在精确标定外参数的情况下，通过最小化视觉特征点的重投影误差进行时间标定，这表明了转台记录的LSLVS轨迹轮廓与从图像中提取的轨迹轮廓之间的异步性。详细分析了系统外参数和时滞的可观性。仿真实验和真实实验表明，该方法具有较高的精度和较强的鲁棒性。
Line structured light vision sensor (LSLVS) with given movement is widely used in many fields of industry for simple structure, fast scanning speed, and low power consumption. One of the applications is rotating LSLVS as a 3-D LiDAR, which is applicable to tasks with limited power consumption, such as the planetary exploration task. Accurate calibration between the LSLVS and the rotating platform is essential to such system, which is difficult for the characteristics of the movement and the problem of time synchronization. In this work, we aim to perform the spatial and the temporal calibration for the rotating LSLVS in two stages. In the first stage, the extrinsic parameters are calculated based on the image data captured in different orientations, during which the LSLVS remains static to avoid errors caused by time offset. Also, multiple constraints are added in this stage in order to have the high accuracy and robustness of calibration. Then, in the second stage, with the extrinsic parameters being precisely calibrated, the temporal calibration is performed by minimizing the reprojection errors of visual feature points, which indicates the asynchronization between the trajectory profile of the LSLVS recorded by the rotating platform and retrieved from images. The observability of the extrinsic parameters and the time delay are analyzed in detail. Simulated and realistic experiments show the competitive accuracy and robustness of the proposed method.