Robust Coarse-to-Fine Registration Scheme for Mobile Laser Scanner Point Clouds Using Multiscale Eigenvalue Statistic-Based Descriptor

Abstract

To overcome the drawbacks of pairwise registration for mobile laser scanner (MLS) point clouds, such as difficulty in searching the corresponding points and inaccuracy registration matrix, a robust coarse-to-fine registration method is proposed to align different frames of MLS point clouds into a common coordinate system. The method identifies the correct corresponding point pairs from the source and target point clouds, and then calculates the transform matrix. First, the performance of a multiscale eigenvalue statistic-based descriptor with different combinations of parameters is evaluated to identify the optimal combination. Second, based on the geometric distribution of points in the neighborhood of the keypoint, a weighted covariance matrix is constructed, by which the multiscale eigenvalues are calculated as the feature description language. Third, the corresponding points between the source and target point clouds are estimated in the feature space, and the incorrect ones are eliminated via a geometric consistency constraint. Finally, the estimated corresponding point pairs are used for coarse registration. The value of coarse registration is regarded as the initial value for the iterative closest point algorithm. Subsequently, the final fine registration result is obtained. The results of the registration experiments with Autonomous Systems Lab (ASL) Datasets show that the proposed method can accurately align MLS point clouds in different frames and outperform the comparative methods.

A novel method for necking detection and measurement in automotive sheet metal components

“…In fact, the model achieves a registration success rate improvement of over 20% in situations with low overlap. Fu et al [32] introduced a novel point cloud registration framework based on depth map matching. They developed a module that utilizes depth map matching to compute a soft correspondence matrix, and introduced a transformer-based method to generate edges for graph construction.…”

Automatic Pairwise Coarse Registration of Terrestrial Point Clouds Using 3D Line Features

“…The registration performance is evaluated in terms of time performance and registration accuracy. The two aspects are commonly adopted by many researchers [15,21,22,44]. The registration accuracy includes rotation error r e and translation error t e , can be calculated by comparing the transformation matrix obtained by the registration method with the ground-truth.…”

“…Numerous methods calculate the transformation parameters by establishing the point correspondences between source and target point clouds using keypoints and their local shape descriptors. First, some keypoints are selected from the original point cloud by the keypoints extraction methods, such as 2.5D scale-invariant feature transform (SIFT) [16], Harris 3D [17], and intrinsic shape signatures (ISS) [18], to accelerate the speed of registration process; Next, the local shape descriptors of the keypoints are calculated, for example, point feature histogram (PFH) [19], fast point feature histogram (FPFH) [20], multiscale eigenvalues statistic (MEVS) [21], feature descriptor vector (FDV) [22], binary shape context (BSC) [6], local voxelized structure (LoVS) [23], and rotational contour signature (RCS) [24]; Third, by comparing the similarity of the local shape descriptors in the feature space, the corresponding points are identified; Finally, the transformation matrix between pairwise point clouds are estimated by the point correspondences. The abovementioned registration strategy is called descriptor-based method, it can well align pairwise point clouds with large overlap, but it is easy to fail when the overlap is small or there are a lot of noises.…”

An Inverse Node Graph-Based Method for the Urban Scene Segmentation of 3D Point Clouds

“…An approach to the rough classification of mobile laser scanning data based on raster image processing techniques was presented by H ůlková [12], which is based on the specified criteria (e.g., in height, size, position) of the objects in the raster image to judge their categories. Another knowledge-based method is to design shape descriptors to find the specific category, which is widely used in the classification and registration process of point clouds [13,14]. A pairwise 3D shape context descriptor was proposed to accurately extract street light poles from nonground points, which was capable of performing partial object matching and retrieval [15].…”