Segmentation Of LiDAR Point Clouds Research Articles

EPointDA: An End-to-End Simulation-to-Real Domain Adaptation Framework for LiDAR Point Cloud Segmentation

Due to its robust and precise distance measurements, LiDAR plays an important role in scene understanding for autonomous driving. Training deep neural networks (DNNs) on LiDAR data requires large-scale point-wise annotations, which are time-consuming and expensive to obtain. Instead, simulation-to-real domain adaptation (SRDA) trains a DNN using unlimited synthetic data with automatically generated labels and transfers the learned model to real scenarios. Existing SRDA methods for LiDAR point cloud segmentation mainly employ a multi-stage pipeline and focus on feature-level alignment. They require prior knowledge of real-world statistics and ignore the pixel-level dropout noise gap and the spatial feature gap between different domains. In this paper, we propose a novel end-to-end framework, named ePointDA, to address the above issues. Specifically, ePointDA consists of three modules: self-supervised dropout noise rendering, statistics-invariant and spatially-adaptive feature alignment, and transferable segmentation learning. The joint optimization enables ePointDA to bridge the domain shift at the pixel-level by explicitly rendering dropout noise for synthetic LiDAR and at the feature-level by spatially aligning the features between different domains, without requiring the real-world statistics. Extensive experiments adapting from synthetic GTA-LiDAR to real KITTI and SemanticKITTI demonstrate the superiority of ePointDA for LiDAR point cloud segmentation.

Multi-Scale Attentive Aggregation for LiDAR Point Cloud Segmentation

Semantic segmentation of LiDAR point clouds has implications in self-driving, robots, and augmented reality, among others. In this paper, we propose a Multi-Scale Attentive Aggregation Network (MSAAN) to achieve the global consistency of point cloud feature representation and super segmentation performance. First, upon a baseline encoder-decoder architecture for point cloud segmentation, namely, RandLA-Net, an attentive skip connection was proposed to replace the commonly used concatenation to balance the encoder and decoder features of the same scales. Second, a channel attentive enhancement module was introduced to the local attention enhancement module to boost the local feature discriminability and aggregate the local channel structure information. Third, we developed a multi-scale feature aggregation method to capture the global structure of a point cloud from both the encoder and the decoder. The experimental results reported that our MSAAN significantly outperformed state-of-the-art methods, i.e., at least 15.3% mIoU improvement for scene-2 of CSPC dataset, 5.2% for scene-5 of CSPC dataset, and 6.6% for Toronto3D dataset.

Semantic Segmentation of LiDAR Point Cloud Based on CAFF-PointNet

Semantic Segmentation of LiDAR Point Cloud Based on CAFF-PointNet

LiDAR point cloud segmentation through scene viewpoint offset

LiDAR point cloud segmentation through scene viewpoint offset

CANOPY LIDAR POINT CLOUD DATA K-MEANS CLUSTERING WATERSHED SEGMENTATION METHOD

Abstract. Airborne laser LiDAR has widely applied in the accurate extraction of single tree canopy for inventory of precision forestry. Due to the over-segmentation phenomenon occurring in the traditional watershed single-wood segmentation, this paper presents a method, called K – means clustering watershed for single tree segmentation. This method consists of four aspects: The first step is to filter the point cloud to eliminate the interference factors such as ground elevation and other factors that interfere with the LiDAR point cloud segmentation; The second step is to optimize the generation of CHM, generate a CMM based on CHM variable window detection, and obtain the treetop position to provide the pixel center position for subsequent K – means cluster segmentation; The third step is to use the K – means clustering algorithm to perform initial cluster segmentation to extract the target pixels of interest. At this time, the local maximum value detected by the variable window in the second step is used as the center pixel of the cluster; In the fourth step, an improved watershed algorithm based on the similarity of 4 neighborhoods is proposed. The improved watershed algorithm is applied to the K – means initial clustering image to segment the target area, and the over-segmentation results are subsequently processed, and the over-segmentation blocks are combined according to certain criteria. Identify the contour of single canopy from the CHM images of the experimental forest data. The experimental results show that the proposed algorithm can effectively solve the over-segmentation problem happening the traditional watershed algorithm. The accuracy of F, R and P parameters can be improved by 7.1%, 11% and 9.8%.

Learning to Optimally Segment Point Clouds

We focus on the problem of class-agnostic instance segmentation of LiDAR point clouds. We propose an approach that combines graph-theoretic search with data-driven learning: it searches over a set of candidate segmentations and returns one where individual segments score well according to a data-driven point-based model of “objectness”. We prove that if we score a segmentation by the worst objectness among its individual segments, there is an efficient algorithm that finds the optimal worst-case segmentation among an exponentially large number of candidate segmentations. We also present an efficient algorithm for the average-case. For evaluation, we repurpose KITTI 3D detection as a segmentation benchmark and empirically demonstrate that our algorithms significantly outperform past bottom-up segmentation approaches and top-down object-based algorithms on segmenting point clouds.

Hyperspectral lidar point cloud segmentation based on geometric and spectral information

Light detection and ranging (lidar) can record a 3D environment as point clouds, which are unstructured and difficult to process efficiently. Point cloud segmentation is an effective technology to solve this problem and plays a significant role in various applications, such as forestry management and 3D building reconstruction. The spectral information from images could improve the segmentation result, but suffers from the varying illumination conditions and the registration problem. New hyperspectral lidar sensor systems can solve these problems, with the capacity to obtain spectral and geometric information simultaneously. The former segmentation on hyperspectral lidar were mainly based on spectral information. The geometric segmentation method widely used by single wavelength lidar was not employed for hyperspectral lidar yet. This study aims to fill this gap by proposing a hyperspectral lidar segmentation method with three stages. First, Connected-Component Labeling (CCL) using the geometric information is employed for base segmentation. Second, the output components of the first stage are split by the spectral difference using Density-Based Spatial Clustering of Applications with Noise (DBSCAN). Third, the components of the second stage are merged based on the spectral similarity using Spectral Angle Match (SAM). Two indoor experimental scenes were setup for validation. We compared the performance of our mothed with that of the 3D and intensity feature based method. The quantitative analysis indicated that, our proposed method improved the point-weighted score by 19.35% and 18.65% in two experimental scenes, respectively. These results showed that the geometric segmentation method for single wavelength lidar could be combined with the spectral information, and contribute to the more effective hyperspectral lidar point cloud segmentation.

Tree Species Classification in Urban Forests based on LiDAR Point Cloud Segmentation and Hyperspectral Metrics Extraction

Tree Species Classification in Urban Forests based on LiDAR Point Cloud Segmentation and Hyperspectral Metrics Extraction

An Optimal Hierarchical Clustering Approach to Mobile LiDAR Point Clouds

This paper proposes a hierarchical clustering approach for the segmentation of mobile LiDAR point clouds. We perform the hierarchical clustering on unorganized point clouds based on a proximity matrix. The dissimilarity measure in the proximity matrix is calculated by the Euclidean distances between clusters and the difference of normal vectors at given points. The main contribution of this paper is that we succeed to optimize the combination of clusters in the hierarchical clustering. The combination is obtained by achieving the matching of a bipartite graph, and optimized by solving the minimum-cost perfect matching. Results show that the proposed optimal hierarchical clustering (OHC) succeeds to achieve the segmentation of multiple individual objects automatically and outperforms the state-of-the-art LiDAR point cloud segmentation approaches.

Segmentation based building detection approach from LiDAR point cloud

Accurate building detection and reconstruction is an important challenge posed to the remote sensing community dealing with LiDAR point cloud. The inherent geometric nature of LiDAR point cloud provides a new dimension to the remote sensing data which can be used to produce accurate 3D building models at relatively less time compared to traditional photogrammetry based 3D reconstruction methods. 3D segmentation is a key step to bring out the implicit geometrical information from the LiDAR point cloud. This research proposes to use open source point cloud library (PCL) for 3D segmentation of LiDAR point cloud and presents a novel histogram based methodology to separate the building clusters from the non building clusters. The proposed methodology has been applied on two different airborne LiDAR datasets acquired over part of urban region around Niagara Falls, Canada and southern Washington, USA. An overall building detection accuracy of 100% and 82% respectively is achieved for the two datasets. The performance of proposed methodology has been compared with the commercially available Terrasolid software. The results show that the buildings detected using open source point cloud library produce comparable results with the buildings detected using commercial software (buildings detection accuracy: 86.3% and 89.2% respectively for the two datasets).

Investigation on the Weighted RANSAC Approaches for Building Roof Plane Segmentation from LiDAR Point Clouds

RANdom SAmple Consensus (RANSAC) is a widely adopted method for LiDAR point cloud segmentation because of its robustness to noise and outliers. However, RANSAC has a tendency to generate false segments consisting of points from several nearly coplanar surfaces. To address this problem, we formulate the weighted RANSAC approach for the purpose of point cloud segmentation. In our proposed solution, the hard threshold voting function which considers both the point-plane distance and the normal vector consistency is transformed into a soft threshold voting function based on two weight functions. To improve weighted RANSAC’s ability to distinguish planes, we designed the weight functions according to the difference in the error distribution between the proper and improper plane hypotheses, based on which an outlier suppression ratio was also defined. Using the ratio, a thorough comparison was conducted between these different weight functions to determine the best performing function. The selected weight function was then compared to the existing weighted RANSAC methods, the original RANSAC, and a representative region growing (RG) method. Experiments with two airborne LiDAR datasets of varying densities show that the various weighted methods can improve the segmentation quality differently, but the dedicated designed weight functions can significantly improve the segmentation accuracy and the topology correctness. Moreover, its robustness is much better when compared to the RG method.

LiDAR Segmentation using Suitable Seed Points for 3D Building Extraction

Abstract. Effective building detection and roof reconstruction has an influential demand over the remote sensing research community. In this paper, we present a new automatic LiDAR point cloud segmentation method using suitable seed points for building detection and roof plane extraction. Firstly, the LiDAR point cloud is separated into "ground" and "non-ground" points based on the analysis of DEM with a height threshold. Each of the non-ground point is marked as coplanar or non-coplanar based on a coplanarity analysis. Commencing from the maximum LiDAR point height towards the minimum, all the LiDAR points on each height level are extracted and separated into several groups based on 2D distance. From each group, lines are extracted and a coplanar point which is the nearest to the midpoint of each line is considered as a seed point. This seed point and its neighbouring points are utilised to generate the plane equation. The plane is grown in a region growing fashion until no new points can be added. A robust rule-based tree removal method is applied subsequently to remove planar segments on trees. Four different rules are applied in this method. Finally, the boundary of each object is extracted from the segmented LiDAR point cloud. The method is evaluated with six different data sets consisting hilly and densely vegetated areas. The experimental results indicate that the proposed method offers a high building detection and roof plane extraction rates while compared to a recently proposed method.

Segmentation of LiDAR point clouds based on similarity measures in multi-dimensional Euclidean Space

The segmentation of LiDAR point cloud is a basic and key step in 3D reconstruction of architecture.Some problems such as under- segmentation or over- segmentation exist in current point cloud segmentation based on boundary,surface or clustering method. In this paper,a point data segmentation method based on similarity measures in multi- dimension Euclidean Space( SMMES) is presented. The main workflow of this method consists of calculating point normal vector,transforming the raw data combined with image features,calculating Euclidean distance in the multi- dimension space,comparing the similarity between the adjacent points,and segmenting the point data. The method proposed in this paper has solved the problem that geometry and spectral features cannot be used in parallel during the point cloud segmentation. In addition,it has the advantages of both geo- metrical segmentation and color- metrical segmentation,and can improve the accuracy of the point cloud segmentation. The segmentation results of the three different methods which are based on geometry features,spectral features and SMMES respectively were compared with each other by using two sets of data,and the experimental results show that the proposed method is significantly feasible and practical.

Rule-based segmentation of LIDAR point cloud for automatic extraction of building roof planes

Abstract. This paper presents a new segmentation technique for LIDAR point cloud data for automatic extraction of building roof planes. Using the ground height from a DEM (Digital Elevation Model), the raw LIDAR points are separated into two groups: ground and nonground points. The ground points are used to generate a "building mask" in which the black areas represent the ground where there are no laser returns below a certain height. The non-ground points are segmented to extract the planar roof segments. First, the building mask is divided into small grid cells. The cells containing the black pixels are clustered such that each cluster represents an individual building or tree. Second, the non-ground points within a cluster are segmented based on their coplanarity and neighbourhood relations. Third, the planar segments are refined using a rule-based procedure that assigns the common points among the planar segments to the appropriate segments. Finally, another rule-based procedure is applied to remove tree planes which are small in size and randomly oriented. Experimental results on the Vaihingen data set show that the proposed method offers high building detection and roof plane extraction rates.

Delineating Individual Trees from Lidar Data: A Comparison of Vector- and Raster-based Segmentation Approaches

Light detection and ranging (lidar) data is increasingly being used for ecosystem monitoring across geographic scales. This work concentrates on delineating individual trees in topographically-complex, mixed conifer forest across the California’s Sierra Nevada. We delineated individual trees using vector data and a 3D lidar point cloud segmentation algorithm, and using raster data with an object-based image analysis (OBIA) of a canopy height model (CHM). The two approaches are compared to each other and to ground reference data. We used high density (9 pulses/m2), discreet lidar data and WorldView-2 imagery to delineate individual trees, and to classify them by species or species types. We also identified a new method to correct artifacts in a high-resolution CHM. Our main focus was to determine the difference between the two types of approaches and to identify the one that produces more realistic results. We compared the delineations via tree detection, tree heights, and the shape of the generated polygons. The tree height agreement was high between the two approaches and the ground data (r2: 0.93–0.96). Tree detection rates increased for more dominant trees (8–100 percent). The two approaches delineated tree boundaries that differed in shape: the lidar-approach produced fewer, more complex, and larger polygons that more closely resembled real forest structure.

Detection of Building in Airborne Laser Scanner Data and Aerial Images

The automatic extraction of the objects from airborne laser scanner data and aerial images has been a topic of research for decades. Airborne laser scanner data are very efficient source for the detection of the buildings. Half of the world population lives in urban/suburban areas, so detailed, accurate and up-to-date building information is of great importance to every resident, government agencies, and private companies.
 The main objective of this paper is to extract the features for the detection of building using airborne laser scanner data and aerial images. To achieve this objective, a method of integration both LiDAR and aerial images has been explored: thus the advantages of both data sets are utilized to derive the buildings with high accuracy. Airborne laser scanner data contains accurate elevation information in high resolution which is very important feature to detect the elevated objects like buildings and the aerial image has spectral information and this spectral information is an appropriate feature to separate buildings from the trees. Planner region growing segmentation of LiDAR point cloud has been performed and normalized digital surface model (nDSM) is obtained by subtracting DTM from the DSM. Integration of the nDSM, aerial images and the segmented polygon features from the LiDAR point cloud has been carried out. The optimal features for the building detection have been extracted from the integration result. Mean height value of the nDSM, Normalized difference vegetation index (NDVI) and the standard deviation of the nDSM are the effective features. The accuracy assessment of the classification results obtained using the calculated attributes was done. Assessment result yielded an accuracy of almost 92 % explaining the features which are extracted by integrating the two data sets was large extent, effective for the automatic detection of the buildings.

Incremental segmentation of lidar point clouds with an octree‐structured voxel space

AbstractLidar (light detection and ranging) data implicitly contains abundant three‐dimensional spatial information. The segmentation of lidar point clouds is the key procedure for transforming implicit spatial information into explicit spatial information. Common criteria used for point cloud segmentation are proximity and coherence of point distribution. An effective segmentation algorithm may apply various steps or combinations of criteria depending on the application. This paper proposes a four‐step segmentation method for lidar point clouds to deliver incremental segmentation results. Segmentation results of each step can provide the fundamental data for the next step. In the first step, the input point cloud is organised into an octree‐structured voxel space, in which the point neighbourhood is established. In the second step, connected voxels which are not empty are grouped to obtain grouped points based on proximity. The third step is a coplanar point segmentation based on both coherence and proximity, which was performed on each point group obtained in the second step. Finally, neighbouring coplanar point groups are merged into “co‐surface” point groups based on the criteria of plane connection and intersection. This scheme enables an incremental retrieval and analysis of a large lidar data‐set. Experimental results demonstrate the effectiveness of the segmentation algorithm in handling both airborne and terrestrial lidar data. It is anticipated that the incremental segmentation results will be useful for object modelling using lidar data.