Iterative Closest Point Variants Research Articles

Reconstruction-Based Hand–Eye Calibration Using Arbitrary Objects

This paper introduces a flexible hand-eye calibration technique for 3D sensor from reconstruction perspective, with no need for specialized and accurate calibration rig. Our intention is to find the hand-eye relation which simultaneously aligns multi-view point clouds of a common scene into the robot base frame, namely simultaneous calibration and reconstruction. To achieve this goal, a novel variant of iterative closest point (ICP) algorithm based on Gauss-Newton method and Lie algebra is proposed, which iteratively transforms multi-view point clouds into robot base frame, estimates point-to-point correspondences between point clouds then refines the hand-eye relation to minimize the Euclidean distance between corresponding points. In addition to the calibration result, it returns a preliminary reconstruction as byproduct. Cases of degeneracy and applicable condition are given and proved. Using arbitrary daily objects with no prior information and real robotic eye-in-hand system, we verify our method feasible and effective.

Coarse Point Cloud Registration Based on Variational Functionals

Point cloud collection forming a 3D scene typically uses information from multiple data scans. The common approach is to register the point cloud pairs consequentially using a variant of the iterative closest point (ICP) algorithm, but most versions of the ICP algorithm only work correctly for a small movement between two point clouds. This makes it difficult to accumulate multiple scans. Global registration algorithms are also known, which theoretically process point clouds at arbitrary initial positions. Recently, a multiparameter variational functional was described and used in the ICP variant to register point clouds at arbitrary initial positions. The disadvantage of this algorithm was the need for manual selection of parameters. In this paper, a modified version of the algorithm with automatic selection of the model parameters is proposed. The proposed algorithm is a fusion of the ICP and RANSAC concepts. Moreover, the algorithm can be parallelized. The performance of the proposed algorithm is compared with that of known global registration algorithms.

Bayesian iterative closest point for mobile robot localization

Accurate localization of a robot in a known environment is a fundamental capability for successfully performing path planning, manipulation, and grasping tasks. Particle filters, also known as Monte Carlo localization (MCL), are a commonly used method to determine the robot’s pose within its environment. For ground robots, noisy wheel odometry readings are typically used as a motion model to predict the vehicle’s location. Such a motion model requires tuning of various parameters based on terrain and robot type. However, such an ego-motion estimation is not always available for all platforms. Scan matching using the iterative closest point (ICP) algorithm is a popular alternative approach, providing ego-motion estimates for localization. Iterative closest point computes a point estimate of the transformation between two poses given point clouds captured at these locations. Being a point estimate method, ICP does not deal with the uncertainties in the scan alignment process, which may arise due to sensor noise, partial overlap, or the existence of multiple solutions. Another challenge for ICP is the high computational cost required to align two large point clouds, limiting its applicability to less dynamic problems. In this paper, we address these challenges by leveraging recent advances in probabilistic inference. Specifically, we first address the run-time issue and propose SGD-ICP, which employs stochastic gradient descent (SGD) to solve the optimization problem of ICP. Next, we leverage SGD-ICP to obtain a distribution over transformations and propose a Markov Chain Monte Carlo method using stochastic gradient Langevin dynamics (SGLD) updates. Our ICP variant, termed Bayesian-ICP, is a full Bayesian solution to the problem. To demonstrate the benefits of Bayesian-ICP for mobile robotic applications, we propose an adaptive motion model employing Bayesian-ICP to produce proposal distributions for Monte Carlo Localization. Experiments using both Kinect and 3D LiDAR data show that our proposed SGD-ICP method achieves the same solution quality as standard ICP while being significantly more efficient. We then demonstrate empirically that Bayesian-ICP can produce accurate distributions over pose transformations and is fast enough for online applications. Finally, using Bayesian-ICP as a motion model alleviates the need to tune the motion model parameters from odometry, resulting in better-calibrated localization uncertainty.

Robust affine registration method using line/surface normals and correntropy criterion

The problem of matching point clouds is an efficient way of registration, which is significant for many research fields including computer vision, machine learning, and robotics. There may be linear or non-linear transformation between point clouds, but determining the affine relation is more challenging among linear cases. Various methods have been presented to overcome this problem in the literature and one of them is the affine variant of the iterative closest point (ICP) algorithm. However, traditional affine ICP variants are highly sensitive to effects such as noises, deformations, and outliers; the least-square metric is substituted with the correntropy criterion to increase the robustness of ICPs to such effects. Correntropy-based robust affine ICPs available in the literature use point-to-point metric to estimate transformation between point clouds. Conversely, in this study, a line/surface normal that examines point-to-curve or point-to-plane distances is employed together with the correntropy criterion for affine point cloud registration problems. First, the maximum correntropy criterion measure is built for line/surface normal conditions. Then, the closed-form solution that maximizes the similarity between point sets is achieved for 2D registration and extended for 3D registration. Finally, the application procedure of the developed robust affine ICP method is given and its registration performance is examined through extensive experiments on 2D and 3D point sets. The results achieved highlight that our method can align point clouds more robustly and precisely than the state-of-the-art methods in the literature, while the registration time of the process remains at reasonable levels.

Real-Time 2-D Lidar Odometry Based on ICP

This study presents a 2-D lidar odometry based on an ICP (iterative closest point) variant used in a simple and straightforward platform that achieves real-time and low-drift performance. With a designated multi-scale feature extraction procedure, the lidar cloud information can be utilized at multiple levels and the speed of data association can be accelerated according to the multi-scale data structure, thereby achieving robust feature extraction and fast scan-matching algorithms. First, on a large scale, the lidar point cloud data are classified according to the curvature into two parts: smooth collection and rough collection. Then, on a small scale, noise and unstable points in the smooth or rough collection are filtered, and edge points and corner points are extracted. Then, the proposed tangent-vector-pairs based on edge and corner points are applied to evaluate the rotation term, which is significant for producing a stable solution in motion estimation. We compare our performance with two excellent open-source SLAM algorithms, Cartographer and Hector SLAM, using collected and open-access datasets in structured indoor environments. The results indicate that our method can achieve better accuracy.

DNSS: Dual-Normal-Space Sampling for 3-D ICP Registration

Rigid registration is a fundamental process in many applications that require alignment of different datasets. Iterative closest point (ICP) is a widely used algorithm that iteratively finds point correspondences and updates the rigid transformation. One of the key variants of ICP to its success is the selection of points, which is directly related to the convergence and robustness of the ICP algorithm. Besides uniform sampling, there are a number of normal-based and feature-based approaches that consider normal, curvature, and/or other signals in the point selection. Among them, normal-space sampling (NSS) is one of the most popular techniques due to its simplicity and low computational cost. The rationale of NSS is to sample enough constraints to determine all the components of transformation, but this paper finds that NSS actually can constrain the translational normal space only. This paper extends the fundamental idea of NSS and proposes Dual NSS (DNSS) to sample points in both translational and rotational normal spaces. Compared with NSS, this approach has similar simplicity and efficiency without any need of additional information, but has a much better effectiveness. Experimental results show that DNSS can outperform the normal-based and feature-based methods in terms of convergence and robustness. For example, DNSS can achieve convergence from an orthogonal initial position while no other methods can achieve. Note to Practitioners-ICP is commonly used to align different data to a same coordination system. While NSS is often used to speed up the alignment process by down-sampling the data uniformly in the normal space. The implementation of NSS only has three steps: 1) construct a set of buckets in the normal-space; 2) put all points of the data into buckets based on their normal direction; and 3) uniformly pick points from all the buckets until the desired number of points is selected. The algorithm is simple and fast, so that it is still the common practice. However, the weakness of NSS comes from the reason that it cannot handle rotational uncertainties. In this paper, a new algorithm called DNSS is developed to constrain both translation and rotation at the same time by introducing a dual-normal space. With a new definition of the normal space, the algorithm complexity of DNSS is the same as that of NSS, and it can be readily implemented in all types of application that are currently using ICP. The experimental results show that DNSS has better efficiency, quality, and reliability than both normal-based and feature-based methods have.

Three-Dimensional Point Cloud Registration Based on Normal Vector Angle

The widespread use of three-dimensional laser scanning makes point cloud registration essential for model reconstruction. Although iterative closest point (ICP) is a widely used algorithm for automatic and accurate point cloud registration, the ICP algorithm is time-intensive and typically falls in local optima. We employ the normal vector angle of the 3D model and integrate a heuristic search into the ICP algorithm to improve its efficiency, especially while registering complex surfaces. We compare this improved ICP algorithm with the original ICP algorithm and variants of the algorithm in terms of convergence, robustness, and response from a poor initial cloud point position to verify that the proposed improvements outperform the conventional ICP algorithm in the three evaluated aspects when processing point clouds with complex surfaces.

3D Reconstruction using artificial bee colony based iterative closest point algorithm

3D Reconstruction has been an enduring problem in Image understanding and computer vision. There is an increasing interest on 3D information in the general public due to rapid development of 3D imaging techniques, marketing of 3D movies and games, and low cost of depth cameras. Numerous algorithms that have been proposed for performing 3D reconstruction using different variants of Iterative Closest Point(ICP) algorithm focusses mainly on reducing the computation time. The accuracy of 3D reconstruction is not taken in to consideration. An efficient, accurate, real time and active 3D reconstruction method using Kinect sensor is developed in this paper focusing on improving the accuracy of 3D reconstruction in less computation time. An Artificial Bee colony based ICP algorithm is proposed by incorporating several efficient variants of ICP algorithm. The proposed algorithm is intended to improve the accuracy and stability of the standard ICP algorithm. The performance of the proposed algorithm is satisfactory when compared with structured light technique and several ICP variants with respect to accuracy, complexity and computation speed.

Three-Dimensional Registration for Handheld Profiling Systems Based on Multiple Shot Structured Light.

In this article, a multi-view registration approach for the 3D handheld profiling system based on the multiple shot structured light technique is proposed. The multi-view registration approach is categorized into coarse registration and point cloud refinement using the iterative closest point (ICP) algorithm. Coarse registration of multiple point clouds was performed using relative orientation and translation parameters estimated via homography-based visual navigation. The proposed system was evaluated using an artificial human skull and a paper box object. For the quantitative evaluation of the accuracy of a single 3D scan, a paper box was reconstructed, and the mean errors in its height and breadth were found to be 9.4 μm and 23 μm, respectively. A comprehensive quantitative evaluation and comparison of proposed algorithm was performed with other variants of ICP. The root mean square error for the ICP algorithm to register a pair of point clouds of the skull object was also found to be less than 1 mm.

Three new Iterative Closest Point variant-methods that improve scan matching for surface mining terrain

The Iterative Closest Point (ICP) algorithm seeks to minimize the misalignment between two point cloud data sets. A limitation of many ICP algorithms is that they work well for some contexts, yet perform poorly in others. Previous work has suggested that the ability of ICP variants to find correspondence was hindered by the presence of geometric disorder in the scene. This paper introduces three new methods based on characterizing the geometric properties of a point using information of its nearest neighbours. Two methods are entropy based and quantify the geometric disorder (eigentropy) in order to improve the filtering of data and thereby remove points that are likely to provide spurious associations. The third method is a point matching method using normals to preferentially work with planar areas of a point cloud. A set of 73,728 ICP variants obtained by combination/permutation of 26 methods are evaluated. These variants were evaluated using a scan matching exercise requiring construction of terrain maps based on data from a mobile sensing platform in an open-cut mining environment. The proposed methods improve ICP performance, as measured by accuracy, precision, and computational efficiency. Notably, five ICP variants, each featuring the new methods of this paper, simultaneously met the solution requirements for three different terrain scenes. It is asserted that being able to characterize the geometric disorder in the point clouds improves the capability of ICP to establish associations between points.

How do ICP variants perform when used for scan matching terrain point clouds?

Many variants of the Iterative Closest Point (ICP) algorithm have been proposed for registering point clouds. This paper explores the performance of 20,736 ICP variants applied to the registration of point clouds for the purpose of terrain mapping, using data obtained from a mobile platform. The methodology of the study has involved taking sequences of 100 consecutive scans at three distinct scenes along the route of a mining haul truck operating in a typical surface mining environment. The scan sequences were obtained at 20 Hz from a Velodyne HDL-64E mounted on the truck. The aim is to understand how well the ICP variants perform in consolidating these scans into sub-maps. Variants are compared against three metrics: accuracy, precision, and relative computational cost. The main finding of the paper is that none of the variants is simultaneously accurate, precise, and fast to compute, across all three scenes. The best performing variants employed strategies that filtered the data sets, used local surface geometry in the form normals, and used the distance between points in one point cloud to a corresponding surface from a reference point cloud as a measure of the fit between two point clouds. The significance of this work is that it: (i) provides guidance in the construction of ICP variants for terrain mapping; and (ii) identifies the significant limitations of existing ICP variants for this application.

A Modified Iterative Closest Point Algorithm for 3D Point Cloud Registration

AbstractIn this article, an accurate method for the registration of point clouds returned by a 3D rangefinder is presented. The method modifies the well‐known iterative closest point (ICP) algorithm by introducing the concept of deletion mask. This term is defined starting from virtual scans of the reconstructed surfaces and using inconsistencies between measurements. In this way, spatial regions of implicit ambiguities, due to edge effects or systematical errors of the rangefinder, are automatically found. Several experiments are performed to compare the proposed method with three ICP variants. Results prove the capability of deletion masks to aid the point cloud registration, lowering the errors of the other ICP variants, regardless the presence of artifacts caused by small changes of the sensor view‐point and changes of the environment.

Intelligent surveillance of indoor environments based on computer vision and 3D point cloud fusion

A real-time detection algorithm for intelligent surveillance is presented. The system, based on 3D change detection with respect to a complex scene model, allows intruder monitoring and detection of added and missing objects, under different illumination conditions.The proposed system has two independent stages. First, a mapping application provides an accurate 3D wide model of the scene, using a view registration approach. This registration is based on computer vision and 3D point cloud. Fusion of visual features with 3D descriptors is used in order to identify corresponding points in two consecutive views. The matching of these two views is first estimated by a pre-alignment stage, based on the tilt movement of the sensor, later they are accurately aligned by an Iterative Closest Point variant (Levenberg–Marquardt ICP), which performance has been improved by a previous filter based on geometrical assumptions.The second stage provides accurate intruder and object detection by means of a 3D change detection approach, based on Octree volumetric representation, followed by a clusters analysis. The whole scene is continuously scanned, and every captured is compared with the corresponding part of the wide model thanks to the previous analysis of the sensor movement parameters. With this purpose a tilt-axis calibration method has been developed.Tests performed show the reliable performance of the system under real conditions and the improvements provided by each stage independently. Moreover, the main goal of this application has been enhanced, for reliable intruder detection by the tilting of the sensors using its built-in motor to increase the size of the monitored area.

Novel automated methods for coarse and fine registrations of point clouds in high precision metrology

Several measuring systems can be combined to perform accurate assessments at the sub-micrometer level in dimensional metrology. The obtained data are fused into a common coordinate system using registration methods for which the optimal transformation parameters from the common parts of the data called correspondences are computed. New original automated coarse and fine registration methods are proposed here using discrete curvatures: an improved Hough transformation method for the coarse registration and three iterative closest point (ICP) variants for the fine registration. The enhancement of Hough consists of exploiting the curvature parameters in order to minimize the basic algorithm complexity. Thus, local transformation parameters are only computed for points presenting similar precalculated surface type. While the rough alignment of the scene data and the model data is thereafter optimized through the fine registration using common ICP algorithm, the first ICP variant includes the curvedness and surface type similarity constraints, especially to reduce the searching area during the matching step. For the proposed second ICP variant, correspondences are searched using a specific distance criterion involving curvature feature similarity measure defined from principal curvatures. The third ICP variant combines both point-to-point and point-to-plane minimizations automatically weighted in the objective function, with the use of moving least squares (MLS) surface technique to determine the corresponding point in point-to-point part. The three developed methods are tested on simulated and real data obtained from a computer tomography (CT) system. The results reveal the benefit of the proposed new automated coarse and fine registration approaches.

Evaluation of Partially Overlapping 3D Point Cloud's Registration by using ICP variant and CloudCompare.

Abstract. Terrestrial Laser Scanners (TLS) are used to get dense point samples of large object’s surface. TLS is new and efficient method to digitize large object or scene. The collected point samples come into different formats and coordinates. Different scans are required to scan large object such as heritage site. Point cloud registration is considered as important task to bring different scans into whole 3D model in one coordinate system. Point clouds can be registered by using one of the three ways or combination of them, Target based, feature extraction, point cloud based. For the present study we have gone through Point Cloud Based registration approach. We have collected partially overlapped 3D Point Cloud data of Department of Computer Science & IT (DCSIT) building located in Dr. Babasaheb Ambedkar Marathwada University, Aurangabad. To get the complete point cloud information of the building we have taken 12 scans, 4 scans for exterior and 8 scans for interior façade data collection. There are various algorithms available in literature, but Iterative Closest Point (ICP) is most dominant algorithms. The various researchers have developed variants of ICP for better registration process. The ICP point cloud registration algorithm is based on the search of pairs of nearest points in a two adjacent scans and calculates the transformation parameters between them, it provides advantage that no artificial target is required for registration process. We studied and implemented three variants Brute Force, KDTree, Partial Matching of ICP algorithm in MATLAB. The result shows that the implemented version of ICP algorithm with its variants gives better result with speed and accuracy of registration as compared with CloudCompare Open Source software.

A new methodology in fast and accurate matching of the 2D and 3D point clouds extracted by laser scanner systems

Abstract Registration of the point clouds is a conventional challenge in computer vision related applications. As an application, matching of train wheel profiles extracted from two viewpoints is studied in this paper. The registration problem is formulated into an optimization problem. An error minimization function for registration of the two partially overlapping point clouds is presented. The error function is defined as the sum of the squared distance between the source points and their corresponding pairs which should be minimized. The corresponding pairs are obtained thorough Iterative Closest Point (ICP) variants. Here, a point-to-plane ICP variant is employed. Principal Component Analysis (PCA) is used to obtain tangent planes. Thus it is shown that minimization of the proposed objective function diminishes point-to-plane ICP variant. We utilized this algorithm to register point clouds of two partially overlapping profiles of wheel train extracted from two viewpoints in 2D. Also, a number of synthetic point clouds and a number of real point clouds in 3D are studied to evaluate the reliability and rate of convergence in our method compared with other registration methods.

Using retinex for point selection in 3D shape registration

Inspired by retinex theory, we propose a novel method for selecting key points from a depth map of a 3D freeform shape; we also use these key points as a basis for shape registration. To find key points, first, depths are transformed using the Hotelling method and normalized to reduce their dependence on a particular viewpoint. Adaptive smoothing is then applied using weights which decrease with spatial gradient and local inhomogeneity; this preserves local features such as edges and corners while ensuring smoothed depths are not reduced. Key points are those with locally maximal depths, faithfully capturing shape. We show how such key points can be used in an efficient registration process, using two state-of-the-art iterative closest point variants. A comparative study with leading alternatives, using real range images, shows that our approach provides informative, expressive, and repeatable points leading to the most accurate registration results.

Accurate registration of MMS point clouds of urban areas using trajectory

Abstract. Recently, by Mobile Mapping System (MMS) with laser scanners, a GPS and IMU (Inertial Measurement Unit), 3D point clouds of urban areas (MMS point clouds) are easily acquired. When the same areas are scanned several times by the MMS, the point clouds often have differences in the range of several hundreds of millimetres. Such differences are caused by inertial drifts of IMU and losses of GPS signals in urban areas. In this paper, we propose an automatic accurate registration method of MMS point clouds using a new variant of ICP (Iterative Closest Point) algorithm for MMS point clouds and trajectory modification. Our method consists of four steps. Firstly, some trajectory points are automatically extracted by analyzing the trajectory. Secondly, the differences of point clouds are derived at the extracted trajectory points in the overlapping scan region by our new ICP algorithm which minimizes pointto- plane and point-to-point distances simultaneously and filters incorrect correspondences based on a point classification by PCA (Principle Component Analysis). Thirdly, the modified positions and rotation parameters at all extracted trajectory points are derived by a least squares method for positioning and registration constraints. Finally, each point in the point clouds is modified by coordinate transformations which are derived from linear interpolation of the modified positions and rotation parameters of the extracted trajectory points. Our method was applied to MMS point clouds and trajectory and the performances were evaluated.

Comparing ICP variants on real-world data sets

Many modern sensors used for mapping produce 3D point clouds, which are typically registered together using the iterative closest point (ICP) algorithm. Because ICP has many variants whose performances depend on the environment and the sensor, hundreds of variations have been published. However, no comparison frameworks are available, leading to an arduous selection of an appropriate variant for particular experimental conditions. The first contribution of this paper consists of a protocol that allows for a comparison between ICP variants, taking into account a broad range of inputs. The second contribution is an open-source ICP library, which is fast enough to be usable in multiple real-world applications, while being modular enough to ease comparison of multiple solutions. This paper presents two examples of these field applications. The last contribution is the comparison of two baseline ICP variants using data sets that cover a rich variety of environments. Besides demonstrating the need for improved ICP methods for natural, unstructured and information-deprived environments, these baseline variants also provide a solid basis to which novel solutions could be compared. The combination of our protocol, software, and baseline results demonstrate convincingly how open-source software can push forward the research in mapping and navigation.

Point set augmentation through fitting for enhanced ICP registration of point clouds in multisensor coordinate metrology

In multisensor coordinate metrology scenarios involving the fusion of homogenous data, specifically 3D point clouds like those originated by CMMs and structured light scanners, the problem of registration, i.e. the proper localization of the clouds in the same coordinate system, is of central importance. For fine registration, known variants of the Iterative Closest Point (ICP) algorithm are commonly adopted; however, no attempt seems to be done to tweak such algorithms to better suit the distinctive multisensor nature of the data. This work investigates an original approach that targets issues which are specific to multisensor coordinate metrology scenarios, such as coexistence of point sets with different densities, different spatial arrangements (e.g. sparse CMM points vs. gridded sets from light scanners), and different noise levels associated to the point sets depending on the metrological performances of the sensors involved. The proposed approach is based on combining known ICP variants with novel point set augmentation techniques, where new points are added to existing sets with the purpose of improving registration performance and robustness to measurement error. In particular, augmentation techniques based on advanced fitting solutions promote a paradigm shift for registration, which is not seen as a geometric problem consisting in moving point sets as close as possible to each other, but as a problem where it is not the original points, but the underlying geometries that must be brought together. In this work, promising combinations of ICP and point augmentation techniques are investigated through the application to virtual scenarios involving synthetic geometries and simulated measurements. Guidelines for approaching registration problems in industrial scenarios involving multisensor data fusion are also provided. Highlights? Point set registration in multisensor coordinate metrology is investigated. ? Iterative Closest Point (ICP) combined with point set augmentation is proposed. ? Point set augmentation improves ICP registration performance. ? Augmentation based on fitting makes ICP more robust to measurement error.