Point Cloud Clustering Research Articles

Automated body measurement of beef cattle based on keypoint detection and local point cloud clustering

Abstract Body size parameters of beef cattle are crucial for assessing growth status and breeding value. In actual farming environments, the various postures of beef cattle and complex backgrounds can affect the accuracy and stability of non-contact body measurement methods. Therefore, this paper proposes a novel method called the cattle body measurement method (CBMM), which combines keypoint detection with local point cloud clustering. First, a keypoint detection model based on YOLOv8-SimBiFPN is constructed. This model enhances the feature extraction and fusion capabilities of YOLOv8-pose by introducing SimAM and BiFPN into the backbone and neck networks, respectively, and realizes 2D keypoint detection for beef cattle in various postures. Second, a 3D keypoint-locating algorithm based on Density-based spatial clustering of applications with noise (DBSCAN) is proposed. This algorithm utilizes 2D keypoints, depth maps and camera parameters to generate local point clouds, which are then clustered using DBSCAN to segment cattle body point clouds, thereby relocating the 3D keypoints based on their positional features. Finally, body size parameters are calculated based on the 3D keypoints and distance formulae. In our experiment, the mean average precision (mAP@0.5) of YOLOv8-SimBiFPN reached 99.1% on an Angus beef cattle keypoint detection dataset. The mean absolute percentage errors for measuring beef cattle withers height, hip height, body depth, body length, and oblique body length using the CBMM were 4.37%, 4.96%, 6.47%, 4.84%, and 4.14%, respectively. In summary, our method can achieve non-contact body measurement for beef cattle in a free-moving state with high accuracy and stability.

Collaborative Obstacle Detection for Dual USVs Using MGNN-DANet with Movable Virtual Nodes and Double Attention

To reduce missed detections in LiDAR-based obstacle detection, this paper proposes a dual unmanned surface vessels (USVs) obstacle detection method using the MGNN-DANet template matching framework. Firstly, point cloud templates for each USV are created, and a clustering algorithm extracts suspected targets from the point clouds captured by a single USV. Secondly, a graph neural network model based on the movable virtual nodes is designed, introducing a neighborhood distribution uniformity metric. This model enhances the local point cloud distribution features of the templates and suspected targets through a local sampling strategy. Furthermore, a feature matching model based on double attention is developed, employing self-attention to aggregate the features of the templates and cross-attention to evaluate the similarity between suspected targets and aggregated templates, thereby identifying and locating another USV within the targets detected by each USV. Finally, the deviation between the measured and true positions of one USV is used to correct the point clouds obtained by the other USV, and obstacle positions are annotated through dual-view point cloud clustering. Experimental results show that, compared to single USV detection methods, the proposed method reduces the missed detection rate of maritime obstacles by 7.88% to 14.69%.

A Small-Object-Detection Algorithm Based on LiDAR Point-Cloud Clustering for Autonomous Vehicles.

3D object-detection based on LiDAR point clouds can help driverless vehicles detect obstacles. However, the existing point-cloud-based object-detection methods are generally ineffective in detecting small objects such as pedestrians and cyclists. Therefore, a small-object-detection algorithm based on clustering is proposed. Firstly, a new segmented ground-point clouds segmentation algorithm is proposed, which filters out the object point clouds according to the heuristic rules and realizes the ground segmentation by multi-region plane-fitting. Then, the small-object point cloud is clustered using an improved DBSCAN clustering algorithm. The K-means++ algorithm for pre-clustering is used, the neighborhood radius is adaptively adjusted according to the distance, and the core point search method of the original algorithm is improved. Finally, the detection of small objects is completed using the directional wraparound box model. After extensive experiments, it was shown that the precision and recall of our proposed ground-segmentation algorithm reached 91.86% and 92.70%, respectively, and the improved DBSCAN clustering algorithm improved the recall of pedestrians and cyclists by 15.89% and 9.50%, respectively. In addition, visualization experiments confirmed that our proposed small-object-detection algorithm based on the point-cloud clustering method can realize the accurate detection of small objects.

Object Detection and Information Perception by Fusing YOLO-SCG and Point Cloud Clustering.

Robots need to sense information about the external environment before moving, which helps them to recognize and understand their surroundings so that they can plan safe and effective paths and avoid obstacles. Conventional algorithms using a single sensor cannot obtain enough information and lack real-time capabilities. To solve these problems, we propose an information perception algorithm with vision as the core and the fusion of LiDAR. Regarding vision, we propose the YOLO-SCG model, which is able to detect objects faster and more accurately. When processing point clouds, we integrate the detection results of vision for local clustering, improving both the processing speed of the point cloud and the detection effectiveness. Experiments verify that our proposed YOLO-SCG algorithm improves accuracy by 4.06% and detection speed by 7.81% compared to YOLOv9, and our algorithm excels in distinguishing different objects in the clustering of point clouds.

Detection-first tightly-coupled LiDAR-Visual-Inertial SLAM in dynamic environments

To address the challenges posed by the dynamic environment for Simultaneous Localization and Mapping (SLAM), a detection-first tightly-coupled LiDAR-Visual-Inertial SLAM incorporating lidar, camera, and inertial measurement unit (IMU) is proposed. Firstly, the point cloud clustering with semantic labels are obtained by fusing image and point cloud information. Then, a tracking algorithm is applied to obtain the information of the motion state of the targets. Afterwards, the tracked dynamic targets are utilized to eliminate extraneous feature points. Finally, a factor graph is used to jointly optimize the IMU pre-integration, and tightly couple the laser odometry and visual odometry within the system. To validate the performance of the proposed SLAM framework, both public datasets (KITTI and UrbanNav) and actual scene data are tested. The experimental results show that compared with LeGO-LOAM, LIO-SAM and LVI-SAM for public dataset, the root mean squared error (RMSE) of proposed algorithm is decreased by 44.56 % (4.47 m) and 4.15 % (4.62 m) in high dynamic scenes and normal scenes, respectively. Through actual scene data, the proposed algorithm mitigates the impact of dynamic objects on map building directly.

Camera LiDAR calibration: an automatic and accurate method with novel PLE metrics

Abstract Camera-LiDAR calibration is a crucial aspect of perception systems and sensor fusion in various fields, facilitating the fusion of data from cameras and LiDAR sensors for applications such as autonomous vehicles, robotics, and augmented reality. In this paper, we presented a novel multi-modal fusion method that introduces an efficient camera-LiDAR joint calibration technique using a simple checkerboard. Our method estimates the 3D rigid transformation of the camera system with respect to the LiDAR frame by establishing 2D-to-3D point correspondences for geometric calibration. We improved the idea of the traditional PNP (Perspective-N-Point) algorithm by employing point cloud segmentation and clustering to re-match the checkerboard. When compared with the calibration methods in the Autoware and MATLAB calibration toolbox, the reprojection error was improved by 38.13\% and 58.30\% respectively. The average translation error was improved by 78.26\% and 51.61\% respectively, while the average rotation error is improved by 75\% and 60.78\% respectively, which has a great improvement in all three areas. At the same time, we propose a new evaluation metric for joint calibration results based on reprojected images: PLE (Pixel-Level Evaluation) index to reflect the accuracy of the joint calibration of the camera and LiDAR. An automatic calibration software has been developed for the calibration of the camera's intrinsic parameters as well as for the joint calibration of the camera-LIDAR extrinsic parameters. We have extensively validated our algorithm using the Intel RealSense Depth D435 Camera and LeiShen C16-151B LiDAR.

The algorithm for denoising point clouds of annular forgings based on Grassmann manifold and density clustering

Abstract In the industrial sector, annular forgings serve as critical load-bearing components in mechanical equipment. During the production process, the precise measurement of the dimensional parameters of annular forgings is of paramount importance to ensure their quality and safety. However, owing to the influence of the measurement environment, the manufacturing process of annular forgings can introduce varying degrees of noise, resulting in inaccurate dimensional measurements. Therefore, researching methods for three-dimensional point cloud data to eliminate noise in annular forging point clouds is of significant importance for improving the accuracy of forging measurements. This paper presents a denoising approach for three-dimensional point cloud data of annular forgings based on Grassmann manifold and density clustering (GDAD). First, within the Grassmann manifold, the core points for density clustering are determined using density parameters. Second, density clustering is performed within the Grassmann manifold, with the Cauchy distance replacing the Euclidean distance to reduce the impact of noise and outliers on the analysis results. Finally, a search tree model was constructed to filter out incorrect point cloud clusters. The fusion of clustering results and the search tree model achieved denoising of point cloud data. Simulation experiments on annular forgings demonstrate that GDAD effectively eliminates edge noise in annular forgings and performs well in denoising point-cloud models with varying levels of noise intensity.

Research on a Train Safety Driving Method Based on Fusion of an Incremental Clustering Algorithm and Lightweight Shared Convolution.

This paper addresses the challenge of detecting unknown or unforeseen obstacles in railway track transportation, proposing an innovative detection strategy that integrates an incremental clustering algorithm with lightweight segmentation techniques. In the detection phase, the paper innovatively employs the incremental clustering algorithm as a core method, combined with dilation and erosion theories, to expand the boundaries of point cloud clusters, merging adjacent point cloud elements into unified clusters. This method effectively identifies and connects spatially adjacent point cloud clusters while efficiently eliminating noise from target object point clouds, thereby achieving more precise recognition of unknown obstacles on the track. Furthermore, the effective integration of this algorithm with lightweight shared convolutional semantic segmentation algorithms enables accurate localization of obstacles. Experimental results using two combined public datasets demonstrate that the obstacle detection average recall rate of the proposed method reaches 90.3%, significantly enhancing system reliability. These findings indicate that the proposed detection strategy effectively improves the accuracy and real-time performance of obstacle recognition, thereby presenting important practical application value for ensuring the safe operation of railway tracks.

PointStaClu: A Deep Point Cloud Clustering Method Based on Stable Cluster Discrimination

Potential inconsistencies between the goals of unsupervised representation learning and clustering within multi-stage deep clustering can diminish the effectiveness of these techniques. However, because the goal of unsupervised representation learning is inherently flexible and can be tailored to clustering, we introduce PointStaClu, a novel single-stage point cloud clustering method. This method employs stable cluster discrimination (StaClu) to tackle the inherent instability present in single-stage deep clustering training. It achieves this by constraining the gradient descent updates for negative instances within the cross-entropy loss function, and by updating the cluster centers using the same loss function. Furthermore, we integrate entropy constraints to regulate the distribution entropy of the dataset, thereby enhancing the cluster allocation. Our framework simplifies the process, employing a single loss function and an encoder for deep point cloud clustering. Extensive experiments on the ModelNet40 and ShapeNet dataset demonstrate that PointStaClu significantly narrows the performance gap between unsupervised point cloud clustering and supervised point cloud classification, presenting a novel approach to point cloud classification tasks.

Recognition and optimisation method of impact deformation patterns based on point cloud and deep clustering: Applied to thin-walled tubes

The recognition and clustering of deformation modes are key to constructing impact deformation constraints for thin-walled structures. This paper transforms the clustering and recognition problem of structural impact deformation modes into a problem of clustering and recognition of point cloud sequences based on pseudo-labels. The effectiveness of the method is assessed, and the experimental results show that the accuracy of the proposed method can reach up to 92.17 % when using a pre-training deep neural network feature extractor, which is not only close to the 98.50 % accuracy of supervised learning classification models but also has a 16.84 % improvement in accuracy compared to the deep clustering method based on K-Means. Under different clustering conditions, the proposed method can effectively classify and recognise samples with similar deformation modes and has the ability to summarise and induce new deformation modes when the number of clusters exceeds the number of manual labels. Furthermore, this paper presents a multi-objective optimisation method for structural crashworthiness under impact deformation constraints based on the NSGA-II algorithm. This method constructs impact deformation constraints using surrogate models and deformation clustering and recognition models. The experimental results show that the proposed method can effectively constrain the generation of the population. It is found that there are a large number of Pareto solutions that do not belong to the expected impact deformation mode under the condition of no deformation mode constraint. In contrast, almost all the obtained Pareto solutions conform to the expected impact deformation mode under the condition of deformation mode constraint. In summary, under the condition of impact deformation constraint, the obtained Pareto solutions can satisfy the crashworthiness requirements while conforming to the expected impact deformation mode.

Real-Time Point Cloud Clustering Algorithm Based on Roadside LiDAR

Real-Time Point Cloud Clustering Algorithm Based on Roadside LiDAR

Machine Learning for the Semi-Automatic 3D Decay Segmentation and Mapping of Heritage Assets

ABSTRACT In Architectural Heritage (AH), diagnostics represent a key topic, related to the assessment of the state of conservation, for a well-structured and integrated knowledge about the artefact. This is of paramount importance for buildings denoted by historic-artistic interest, often in widespread damaged conditions. Digital technologies could represent a valuable alternative, for the creation, processing, and analysis of 2D images or reality-capture 3D data. Indeed, image processing and artificial intelligence are spreading in the AH domain, for both documentation and conservation purposes. The present research proposes a novel methodological approach, based on the application of machine learning systems, for the assessment and mapping of decay evidence in an ancient artefact, within the diagnostic process. An unsupervised hierarchical clustering of point clouds has been validated and compared against a state-of-the-art supervised random forest. The presented pipeline has been tested on a plurality of case studies, representative of the territorial historic built heritage, qualified by a great variety, in terms of age, building type, surface, and lighting conditions. For most of the classes, the cloud-based unsupervised approach consents to accomplish better performances.

Learnt representations of proteins can be used for accurate prediction of small molecule binding sites on experimentally determined and predicted protein structures

Protein-ligand binding site prediction is a useful tool for understanding the functional behaviour and potential drug-target interactions of a novel protein of interest. However, most binding site prediction methods are tested by providing crystallised ligand-bound (holo) structures as input. This testing regime is insufficient to understand the performance on novel protein targets where experimental structures are not available. An alternative option is to provide computationally predicted protein structures, but this is not commonly tested. However, due to the training data used, computationally-predicted protein structures tend to be extremely accurate, and are often biased toward a holo conformation. In this study we describe and benchmark IF-SitePred, a protein-ligand binding site prediction method which is based on the labelling of ESM-IF1 protein language model embeddings combined with point cloud annotation and clustering. We show that not only is IF-SitePred competitive with state-of-the-art methods when predicting binding sites on experimental structures, but it performs better on proxies for novel proteins where low accuracy has been simulated by molecular dynamics. Finally, IF-SitePred outperforms other methods if ensembles of predicted protein structures are generated.

Omni-OTPE: Omnidirectional Optimal Real-Time Ground Target Position Estimation System for Moving Lightweight Unmanned Aerial Vehicle.

Ground target detection and positioning systems based on lightweight unmanned aerial vehicles (UAVs) are increasing in value for aerial reconnaissance and surveillance. However, the current method for estimating the target's position is limited by the field of view angle, rendering it challenging to fulfill the demands of a real-time omnidirectional reconnaissance operation. To address this issue, we propose an Omnidirectional Optimal Real-Time Ground Target Position Estimation System (Omni-OTPE) that utilizes a fisheye camera and LiDAR sensors. The object of interest is first identified in the fisheye image, and then, the image-based target position is obtained by solving using the fisheye projection model and the target center extraction algorithm based on the detected edge information. Next, the LiDAR's real-time point cloud data are filtered based on position-direction constraints using the image-based target position information. This step allows for the determination of point cloud clusters that are relevant to the characterization of the target's position information. Finally, the target positions obtained from the two methods are fused using an optimal Kalman fuser to obtain the optimal target position information. In order to evaluate the positioning accuracy, we designed a hardware and software setup, mounted on a lightweight UAV, and tested it in a real scenario. The experimental results validate that our method exhibits significant advantages over traditional methods and achieves a real-time high-performance ground target position estimation function.

Research on Individual Tree Canopy Segmentation of Camellia oleifera Based on a UAV-LiDAR System

In consideration of the limited accuracy of individual tree canopy segmentation algorithms due to the diverse canopy structure and complex environments in mountainous and hilly areas, this study optimized the segmentation parameters of three algorithms for individual tree canopy segmentation of Camellia oleifera in such environments by analyzing their respective parameters. Utilizing an Unmanned Aerial Vehicle-Light Detecting and Ranging (UAV-LiDAR) system, we obtained Canopy Height Models (CHM) of Camellia oleifera canopies based on Digital Elevation Models (DEM) and Digital Surface Models (DSM). Subsequently, we investigated the effects of CHM segmentation, point cloud clustering segmentation, and layer stacking fitting segmentation on Camellia oleifera canopies across different research areas. Additionally, combining ground survey data from forest lands with visual interpretation of UAV orthophoto images, we evaluated the performance of these three segmentation algorithms in terms of the F-score as an evaluation indicator for individual tree canopy segmentation accuracy. Combined with the Cloth Simulation Filter (CSF) filtering algorithm after removing the ground point cloud, our findings indicate that among different camellia densities and terrain environments, the point cloud clustering segmentation algorithm achieved the highest segmentation accuracy at 93%, followed by CHM segmentation at 88% and the layer stacking fitting segmentation method at 84%. By analyzing the data from UAV-LiDAR technology involving various land and Camellia oleifera planting types, we verified the applicability of these three segmentation algorithms for extracting camellia canopies. In conclusion, this study holds significant importance for accurately delineating camellia canopies within mountainous hilly environments while providing valuable insights for further research in related fields.

LiDAR Dynamic Target Detection Based on Multidimensional Features.

To address the limitations of LiDAR dynamic target detection methods, which often require heuristic thresholding, indirect computational assistance, supplementary sensor data, or postdetection, we propose an innovative method based on multidimensional features. Using the differences between the positions and geometric structures of point cloud clusters scanned by the same target in adjacent frame point clouds, the motion states of the point cloud clusters are comprehensively evaluated. To enable the automatic precision pairing of point cloud clusters from adjacent frames of the same target, a double registration algorithm is proposed for point cloud cluster centroids. The iterative closest point (ICP) algorithm is employed for approximate interframe pose estimation during coarse registration. The random sample consensus (RANSAC) and four-parameter transformation algorithms are employed to obtain precise interframe pose relations during fine registration. These processes standardize the coordinate systems of adjacent point clouds and facilitate the association of point cloud clusters from the same target. Based on the paired point cloud cluster, a classification feature system is used to construct the XGBoost decision tree. To enhance the XGBoost training efficiency, a Spearman's rank correlation coefficient-bidirectional search for a dimensionality reduction algorithm is proposed to expedite the optimal classification feature subset construction. After preliminary outcomes are generated by XGBoost, a double Boyer-Moore voting-sliding window algorithm is proposed to refine the final LiDAR dynamic target detection accuracy. To validate the efficacy and efficiency of our method in LiDAR dynamic target detection, an experimental platform is established. Real-world data are collected and pertinent experiments are designed. The experimental results illustrate the soundness of our method. The LiDAR dynamic target correct detection rate is 92.41%, the static target error detection rate is 1.43%, and the detection efficiency is 0.0299 s. Our method exhibits notable advantages over open-source comparative methods, achieving highly efficient and precise LiDAR dynamic target detection.

Indoor Radar Point Cloud Ghost Elimination Method Based on Point Cloud Cluster Occlusion

Indoor Radar Point Cloud Ghost Elimination Method Based on Point Cloud Cluster Occlusion

Identification and localization of transmission lines for live working in distribution network

In this paper, aiming at the current distribution network with power operation in the presence of outdoor bright light, the operating scene is complex and not fixed, the transmission lines and lead samples are small and uneven, leading to poor accuracy of machine learning and other methods of distribution network with power operation robot identification and localization, this paper selects three-dimensional LIDAR to design a set of support vector machine (SVM) classifier and correlation analysis based on the target recognition method. This paper selects three-dimensional LIDAR to design a set of support vector machine (SVM) classifier and correlation analysis based on the target recognition method. First, in the context of power distribution network operations, we have explored various filtering methods suitable for preprocessing transmission line data. These methods aim to remove outliers and clutter from point clouds, thus enhancing both point cloud quality and processing speed. This, in turn, improves the efficiency of real-time on-site detection. We conducted experimental comparisons of various clustering algorithms and opted for a region-based point cloud clustering algorithm to achieve the segmentation of individual point clouds. Secondly, we proposed a multi-feature composite approach based on the Viewpoint Feature Histogram (VFH) features, which helps maintain the scale-invariant characteristics of the features. We then introduced a multi-feature composite criterion based on VFH features to extract features from segmented point clouds. Subsequently, an SVM classifier based on this composite feature criterion was developed to achieve target identification and classification. Experimental results have demonstrated improved accuracy in target identification. Finally, we integrated this system with a bucket-arm vehicle for coordinate conversion and precise navigation of the robot to the target station. This method significantly improves the overall operational efficiency of power distribution tasks.

An optimized fuzzy K-means clustering method for automated rock discontinuities extraction from point clouds

Recently, non-contact measurement methods such as laser scanning, have gained popularity in collecting discontinuous data due to their ability to generate high-resolution point clouds containing detailed information about rock surface. However, quickly and accurately extracting discontinuities from massive point clouds faces challenges. In this study, we propose an optimization algorithm based on fuzzy clustering and region growth that enables swift extraction of discontinuity information from point clouds. The proposed method employed a composite indicator to evaluate the similarity and dissimilarity of the points in a clustering group basing on the membership function matrix and the optimal clustering number could be estimated without predefining. Additionally, region growing is difficult to deal with increasingly enormous point clouds, a faster way is estimating a possible range as a search radius to avoid meaningless time-consuming in region growing. Further, the proposed methodology was implemented in Matlab to extract discontinuities from high-resolution point cloud, includes data pre-processing, optimized fuzzy clustering, and optimized region growing. Finally, particular attention was given to the sensitivity of automatic extraction in point cloud resolution and cluster number, these parameters are always special in different objects. The results showed that the optimized method performed excellent in clustering without a priori assumption of cluster number, and provided optional range of resolutions without losses of accuracy to cater to diverse requirements. The proposed method gave a new way to estimate the optimal clustering number as same as manually separated without predefining, to collect all orientations of discontinuity quickly, and to meet different needs with appropriate resolution.

Comparison of two novel methods for counting wheat ears in the field with terrestrial LiDAR

BackgroundThe metrics for assessing the yield of crops in the field include the number of ears per unit area, the grain number per ear, and the thousand-grain weight. Typically, the ear number per unit area contributes the most to the yield. However, calculation of the ear number tends to rely on traditional manual counting, which is inefficient, labour intensive, inaccurate, and lacking in objectivity. In this study, two novel extraction algorithms for the estimation of the wheat ear number were developed based on the use of terrestrial laser scanning (TLS) in conjunction with the density-based spatial clustering (DBSC) algorithm based on the normal and the voxel-based regional growth (VBRG) algorithm. The DBSC involves two steps: (1) segmentation of the point clouds using differences in the normal vectors and (2) clustering of the segmented point clouds using a density clustering algorithm to calculate the ear number. The VBRG involves three steps: (1) voxelization of the point clouds, (2) construction of the topological relationships between the voxels as a connected region using the k-dimensional tree, and (3) detection of the wheat ears in the connected areas using a regional growth algorithm.ResultsThe results demonstrated that DBSC and VBRG were promising in estimating the number of ears for different cultivars, planting densities, N fertilization rates, and growth stages of wheat (RMSE = 76 ~ 114 ears/m2, rRMSE = 18.62 ~ 27.96%, r = 0.76 ~ 0.84). Comparing the performance of the two algorithms, the overall accuracy of the DBSC (RMSE = 76 ears/m2, rRMSE = 18.62%, r = 0.84) was better than that of the VBRG (RMSE = 114 ears/m2, rRMSE = 27.96%, r = 0.76). It was found that with the DBSC, the calculation in points as units permitted more detailed information to be retained, and this method was more suitable for estimation of the wheat ear number in the field.ConclusionsThe algorithms adopted in this study provide new approaches for non-destructive measurement and efficient acquisition of the ear number in the assessment of the wheat yield phenotype.