Pair Of Point Clouds Research Articles

Learning 3D local surface descriptor for point cloud images of objects in the real-world

Surface descriptors, which represent the surface characteristics of an image numerically, are the fundamental elements in many vision applications. Although traditional surface descriptors that are handcrafted or learned using machine learning techniques have been applied in many different vision applications, some difficulty remains in handling large amounts of noise and variance in 3D data. To resolve this difficulty, recent studies have applied deep learning techniques for the development of surface descriptors. Unlike other techniques based on the complete 3D CAD model or pre-known mesh information of the object, we consider the constraint of the robotic applications in which the information mentioned above is difficult to preload. In this paper, we propose a new 3D surface descriptor that does not require any pre-loaded topological information of the objects or a mesh construction, which may occasionally fail with new or previously unknown objects. Further, we propose a voxel representation that is adaptive to the density of the points, resolving the problem of varying densities of the point cloud data. Finally, we adopt domain-adversarial learning that leads a network to learn the features discriminative for similarity measurements while remaining invariant to different point densities. We gathered approximately 5,000 point-cloud images of objects along with their position and orientation information. We then constructed approximately half a million pairs of point clouds indicating the identical and different parts of the objects, which are labeled as true and false, respectively. The dataset of constructed pairs was used for the learning of 3D surface descriptors using a Siamese convolutional neural network (SCNN) with a domain-adversarial characteristic. The results indicate that the proposed descriptor outperforms other descriptors.

COMPARISON WITH ACCURACY OF TERRESTRIAL LASER SCANNER BY USING POINT CLOUD ALIGNED WITH SHAPE MATCHING AND BEST FITTING METHODS

Abstract. Recently operation systems of laser scanning have been obviously improved; for instance shape matching has been equipped with software on a post processing stage so measurement without any targets is a prerequisite condition of field surveying with laser scanners. Moreover a shape matching method enables us to easily register a pair of point clouds with some errors even if those data are scanned by several type scanners. Those slightly errors can influence accuracy of alignments if the object is large to require a lot of scans. Laser scanning data has random errors and accuracy of alignments can be improved by matching error distributions of pairs of point clouds to natural distributions. This method is called “best fitting” in contrast “shape matching” in a software, PolyWorks |Inspector. In this paper, accuracy of alignments between shape matching and best fitting is discussed. The scan data of three phaseshift laser scanners (FARO Focus 3D MS120, FARO Focus 3D X330 and Z+F Imager 5016) and two time-of-flight scanners (Leica BLK 360 and Leica Scan station C5) are used for analyses. Accuracy of alignments by using shape matching and best fitting methods is demonstrated by showing points of scan data with histograms of error distributions.

A Novel Approach to 3-D Change Detection in Multitemporal LiDAR Data Acquired in Forest Areas

Light Detection and Ranging (LiDAR) data have been widely used to characterize the 3-D structure of the forest. However, their use in a multitemporal framework has been quite limited due to the relevant challenges introduced by the comparison of pairs of point clouds. Because of the irregular sampling of the laser scanner and the complex structure of forest areas, it is not possible to perform a point-to-point comparison between the two data. To overcome these challenges, a novel hierarchical approach to the detection of 3-D changes in forest areas is proposed. The method first detects the large changes (e.g., cut trees) by comparing the Canopy Height Models derived from the two LiDAR data. Then, according to an object-based change detection approach, it identifies the single-tree changes by monitoring both the treetop and the crown volume growth. The proposed approach can compare LiDAR data with significantly different pulse densities, thus allowing the use of many data available in real applications. Experimental results pointed out that the method can accurately detect large changes, exhibiting a low rate of false and missed alarms. Moreover, it can detect changes in terms of single-tree growth, which are consistent with the expected growth rates of the considered areas.

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.

COMPARISON OF 2D AND 3D APPROACHES FOR THE ALIGNMENT OF UAV AND LIDAR POINT CLOUDS

Abstract. The automatic alignment of 3D point clouds acquired or generated from different sensors is a challenging problem. The objective of the alignment is to estimate the 3D similarity transformation parameters, including a global scale factor, 3 rotations and 3 translations. To do so, corresponding anchor features are required in both data sets. There are two main types of alignment: i) Coarse alignment and ii) Refined Alignment. Coarse alignment issues include lack of any prior knowledge of the respective coordinate systems for a source and target point cloud pair and the difficulty to extract and match corresponding control features (e.g., points, lines or planes) co-located on both point cloud pairs to be aligned. With the increasing use of UAVs, there is a need to automatically co-register their generated point cloud-based digital surface models with those from other data acquisition systems such as terrestrial or airborne lidar point clouds. This works presents a comparative study of two independent feature matching techniques for addressing 3D conformal point cloud alignment of UAV and lidar data in different 3D coordinate systems without any prior knowledge of the seven transformation parameters.

A WEIGHTED CLOSED-FORM SOLUTION FOR RGB-D DATA REGISTRATION

Existing 3D indoor mapping of RGB-D data are prominently point-based and feature-based methods. In most cases iterative closest point (ICP) and its variants are generally used for pairwise registration process. Considering that the ICP algorithm requires an relatively accurate initial transformation and high overlap a weighted closed-form solution for RGB-D data registration is proposed. In this solution, we weighted and normalized the 3D points based on the theoretical random errors and the dual-number quaternions are used to represent the 3D rigid body motion. Basically, dual-number quaternions provide a closed-form solution by minimizing a cost function. The most important advantage of the closed-form solution is that it provides the optimal transformation in one-step, it does not need to calculate good initial estimates and expressively decreases the demand for computer resources in contrast to the iterative method. Basically, first our method exploits RGB information. We employed a scale invariant feature transformation (SIFT) for extracting, detecting, and matching features. It is able to detect and describe local features that are invariant to scaling and rotation. To detect and filter outliers, we used random sample consensus (RANSAC) algorithm, jointly with an statistical dispersion called interquartile range (IQR). After, a new RGB-D loop-closure solution is implemented based on the volumetric information between pair of point clouds and the dispersion of the random errors. The loop-closure consists to recognize when the sensor revisits some region. Finally, a globally consistent map is created to minimize the registration errors via a graph-based optimization. The effectiveness of the proposed method is demonstrated with a Kinect dataset. The experimental results show that the proposed method can properly map the indoor environment with an absolute accuracy around 1.5% of the travel of a trajectory.

A WEIGHTED CLOSED-FORM SOLUTION FOR RGB-D DATA REGISTRATION

Abstract. Existing 3D indoor mapping of RGB-D data are prominently point-based and feature-based methods. In most cases iterative closest point (ICP) and its variants are generally used for pairwise registration process. Considering that the ICP algorithm requires an relatively accurate initial transformation and high overlap a weighted closed-form solution for RGB-D data registration is proposed. In this solution, we weighted and normalized the 3D points based on the theoretical random errors and the dual-number quaternions are used to represent the 3D rigid body motion. Basically, dual-number quaternions provide a closed-form solution by minimizing a cost function. The most important advantage of the closed-form solution is that it provides the optimal transformation in one-step, it does not need to calculate good initial estimates and expressively decreases the demand for computer resources in contrast to the iterative method. Basically, first our method exploits RGB information. We employed a scale invariant feature transformation (SIFT) for extracting, detecting, and matching features. It is able to detect and describe local features that are invariant to scaling and rotation. To detect and filter outliers, we used random sample consensus (RANSAC) algorithm, jointly with an statistical dispersion called interquartile range (IQR). After, a new RGB-D loop-closure solution is implemented based on the volumetric information between pair of point clouds and the dispersion of the random errors. The loop-closure consists to recognize when the sensor revisits some region. Finally, a globally consistent map is created to minimize the registration errors via a graph-based optimization. The effectiveness of the proposed method is demonstrated with a Kinect dataset. The experimental results show that the proposed method can properly map the indoor environment with an absolute accuracy around 1.5% of the travel of a trajectory.

Automatic Point Clouds Registration Method Based on Mesh Segmentation

This paper proposes an automatic point clouds registration method based on High-Speed Mesh Segmentation. The proposed method works fast for doing an initial registration and extracting point clouds region feature. First, the features of the point region are used for matching point cloud regions. Second, matched regions sets are classified for calculating transform matrix of initial registration. Based on the initial registration result the Iterative Closest Point (ICP) algorithm which had been used for accuracy registration to composite point cloud pairs will be applied. The proposed registration approach is able to do automatic registration without any assumptions about their initial positions, and avoid the problems of traditional ICP in bad initial estimate. The proposed method plus with ICP algorithm provides an efficient 3D model for computer-aided engineering and computer-aided design.

Learning to close loops from range data

In this paper we address the loop closure detection problem in simultaneous localization and mapping ( slam ), and present a method for solving the problem using pairwise comparison of point clouds in both two and three dimensions. The point clouds are mathematically described using features that capture important geometric and statistical properties. The features are used as input to the machine learning algorithm AdaBoost, which is used to build a non-linear classifier capable of detecting loop closure from pairs of point clouds. Vantage point dependency in the detection process is eliminated by only using rotation invariant features, thus loop closure can be detected from an arbitrary direction. The classifier is evaluated using publicly available data, and is shown to generalize well between environments. Detection rates of 66%, 63% and 53% for 0% false alarm rate are achieved for 2D outdoor data, 3D outdoor data and 3D indoor data, respectively. In both two and three dimensions, experiments are performed using publicly available data, showing that the proposed algorithm compares favourably with related work.

Registration of point clouds using sample-sphere and adaptive distance restriction

Registration of point clouds is a fundamental problem in shape acquisition and shape modeling. In this paper, a novel technique, the sample-sphere method, is proposed to register a pair of point clouds in arbitrary initial positions. This method roughly aligns point clouds by matching pairs of triplets of points, which are approximately congruent under rigid transformation. For a given triplet of points, this method can find all its approximately congruent triplets in O(knlog n) time, where n is the number of points in the point cloud, and k is a constant depending only on a given tolerance to the rotation error. By employing the techniques of wide bases and largest common point set (LCP), our method is resilient to noise and outliers. Another contribution of this paper is proposing an adaptive distance restriction to improve ICP (iterative closest point) algorithm, which is a classical method to refine rough alignments. With this restriction, the improved ICP is able to reject unreasonable corresponding point pairs during each iteration, so it can precisely align the point clouds which have large non-overlapping regions.