Semantic-assisted 3D normal distributions transform for scan registration in environments with limited structure

Abstract

Abstract-Point cloud registration is a core problem of many robotic applications, including simultaneous localization and mapping. The Normal Distributions Transform (NDT) is a method that fits a number of Gaussian distributions to the data points, and then uses this transform as an approximation of the real data, registering a relatively small number of distributions as opposed to the full point cloud. This approach contributes to NDT's registration robustness and speed but leaves room for improvement in environments of limited structure.To address this limitation we propose a method for the introduction of semantic information extracted from the point clouds into the registration process. The paper presents a large scale experimental evaluation of the algorithm against NDT on two publicly available benchmark data sets. For the purpose of this test a measure of smoothness is used for the semantic partitioning of the point clouds. The results indicate that the proposed method improves the accuracy, robustness and speed of NDT registration, especially in unstructured environments, making NDT suitable for a wider range of applications.

Integrating Deep Semantic Segmentation Into 3-D Point Cloud Registration

“…Semantic-assisted Normal Distributions Transform registration (SE-NDT) [4], a method we introduced in our previous work, uses semantic information to segment the point cloud and constructs the Normal Distributions Transform for each segment independently. Distribution-to-distribution correspondences are then only searched among the NDTs constructed from the same type of semantic label.…”

“…Semantic-assisted NDT Semantic-assisted Normal Distributions Transform (SE-NDT) [4] is a registration method where the point cloud is partitioned according to per-point semantic labels, and sets of Normal Distributions Transforms are estimated and registered for each partition independently. The algorithm assumes that every point in the point cloud has one semantic label.…”

“…A Gaussian distribution of points is assumed for each voxel. For each distribution the mean vector µ i and covariance matrix C i are estimated according to Equations 3,4. The probability density function at voxel i is then given by Equation 5.…”

“…For FGR, the neighbours for normal estimation are set to 200, the neighbours for FPFH are set to 240 and the number of tuples is limited to 500. The parameters for NDT are taken from our previous work [4]. 1 https://github.com/IntelVCL/FastGlobalRegistration VII.…”

“…In our previous work, we proposed the Semantic-assisted Normal Distributions Transform (SE-NDT) [4], a registration method that uses normal distribution transforms but also considers per-point semantic information that might be available to aid the registration process. An increasing number of robotic systems employ semantic segmentation algorithms for various purposes, for example, to detect traversable terrain, buildings, trees, crops, etc.…”