There is an increasingly urgent need for model attitude measurement technology in fields of urban modeling, aerospace, autonomous driving, etc. Among them, the point cloud registration algorithm is essential. The existing registration algorithms cannot simultaneously meet the requirements of high speed, high precision and large field of view. To this end, this paper proposes a registration algorithm, which combines normal distribution transform (NDT) and iterative closest point (ICP) to perform secondary registration on point clouds. An attitude measurement platform has been built and the LiDAR continuously obtains the point cloud data of the model. Multi-layered voxel and quasi-Newton method were used to accelerate the NDT algorithm, which is used to calculate the transformation matrix of adjacent frames. If the attitude change exceeds the threshold, ICP will be used with the initial solution from NDT. This method combines the high precision of ICP with the high speed of NDT, so that the dynamic model attitude measurement can be in real time under the premise of high precision, and it is suitable for a variety of attitude measurement scenarios with high precision, high speed and large field of view.
|