基于结构光相机的钢筋骨架整体点云获取算法

为了获取钢筋骨架质量自动检查所需的高精度钢筋骨架整体点云，建立了基于结构光相机的钢筋骨架整体点云获取算法。首先，对结构光相机采集得到的多幅钢筋骨架图像进行三维重建，得到结构光相机的无量纲位姿。其次，根据无量纲位姿获取有量纲位姿。然后，计算这些有量纲位姿间精确的转换矩阵。接着，基于这些有量纲位姿及其两两之间的精确转换矩阵，使用图优化对这些有量纲位姿进行优化，以得到高精度的有量纲位姿。最后，基于高精度的有量纲位姿对齐结构光相机采集的所有点云，获取钢筋骨架整体点云。实验结果表明，该算法获取实际预制钢筋混凝土构件钢筋骨架整体点云的耗时约 10 min，且点云的误差约为 5 mm。钢筋骨架整体点云获取算法可以快速获取钢筋骨架整体点云，而且所得点云的精度较高，可以满足钢筋骨架质量自动检查的要求。

A method for obtaining the complete point cloud of reinforcement skeletonsbased on a structured light camera

In order to obtain the high-precision complete point cloud of a reinforcement skeleton required for the automatic quality inspection, an algorithm was proposed for obtaining the complete point cloud of reinforcement skeletons based on a structured light camera. Firstly, 3D reconstruction was carried out for multiple reinforcement skeleton images collected using a structured light camera, thus obtaining the dimensionless poses of the structured light camera. Secondly, the dimensional poses were obtained according to the dimensionless poses. Then, the precise transformation matrix between these dimensional poses was calculated. Next, based on these dimensional poses and the precise transformation matrix between them, graph optimization was employed to optimize these dimensional poses to obtain those with high precision. Finally, point clouds obtained using the structured light camera were aligned based on the dimensional poses with high precision, which can generate the complete point cloud of the reinforcement skeleton. The experimental results show that it would take the proposed algorithm about 10 minutes to obtain the complete point cloud of the reinforcement skeleton of a practical precast concrete component, and the error of the point cloud is around 5 mm. It is concluded that the proposed algorithm can quickly obtain the complete point cloud of the reinforcement skeleton, with high accuracy, which can meet the requirement of the automatic quality inspection of the reinforcement skeleton.