Beyond points: Evaluating recent 3D scan-matching algorithms

Abstract

Beyond points: Evaluating recent 3D scan-matching algorithms. (pp. 3631-3637 Abstract-Given that 3D scan matching is such a central part of the perception pipeline for robots, thorough and large-scale investigations of scan matching performance are still surprisingly few. A crucial part of the scientific method is to perform experiments that can be replicated by other researchers in order to compare different results. In light of this fact, this paper presents a thorough comparison of 3D scan registration algorithms using a recently published benchmark protocol which makes use of a publicly available challenging data set that covers a wide range of environments. In particular, we evaluate two types of recent 3D registration algorithms -one local and one global. Both approaches take local surface structure into account, rather than matching individual points. After well over 100 000 individual tests, we conclude that algorithms using the normal distributions transform (NDT) provides accurate results compared to a modern implementation of the iterative closest point (ICP) method, when faced with scan data that has little overlap and weak geometric structure. We also demonstrate that the minimally uncertain maximum consensus (MUMC) algorithm provides accurate results in structured environments without needing an initial guess, and that it provides useful measures to detect whether it has succeeded or not. We also propose two amendments to the experimental protocol, in order to provide more valuable results in future implementations. In: 2015 IEEE International Conference on Robotics and Automation (ICRA)

Integrating Deep Semantic Segmentation Into 3-D Point Cloud Registration

“…The distribution-to-distribution variant of NDT [13], [18] consists of the following steps to register a point cloud M to a point cloud F. At its first step, the algorithm discretizes the space into voxels. Let S i be the set of points v of F in voxel i.…”

Semi-supervised 3D place categorisation by descriptor clustering

“…NDT has previously been shown to be useful for applications of point cloud registration [9] and many other tasks. One can construct a histogram of NDT voxels to encode point cloud appearance, by classifying the PDFs based on orientation and shape, and distance from the sensor.…”

“…The selection of the farrange cut-off distance relates to the question of "what is a place." Especially with outdoor lidar data, areas that are [3,6), [5,8), [7,10), [9,12), [11,15)[14, ∞)} ranges, outdoor { [5,8), [7,10), [9,12), [11,15), [14,20)} far from the robot position also influence the appearance descriptor, which may not be desired. Nevertheless, the NDT histogram appearance descriptor is remarkably robust to changing the parameters listed in Tab.…”

Precipitation downscaling using a probability-matching approach and geostationary infrared data: An evaluation over six climate regions

“…1c). Magnusson et al (2015) demonstrated that CDF matching works better than histogram-20 matching method when low values have high frequencies, which is generally the case for precipitation.…”