Processing Laser Point Cloud in Fully Mechanized Mining Face Based on DGCNN

MaskNet++: Inlier/outlier identification for two point clouds

A point cloud deep neural network metamodel method for aerodynamic prediction

In a point cloud registration task, the feature extraction ability of the network directly affects the final registration effect. Previous work has tended to ignore the local information of point clouds and the context information between two point clouds in the feature extraction. To overcome the poor feature extraction ability and lack of robustness of existing registration algorithms, this paper, inspired by the graph attention network, proposes a learning-based point cloud registration framework that uses the existing network to extract point cloud features. We call it CPCRNet. Furthermore, a Capture module based on graph attention (GAT) is proposed to re-integrate the point cloud context information. Experimental results indicate that Capture, the core feature extraction module in the network, has a significant impact on the registration effect, with a mean square error, root mean square error, and mean absolute error between the ground truth and predicted values lower than those of competing methods. Moreover, experimental results for ModelNet40 and ShapeNet Part datasets show that the errors are not significantly changed by noise, indicating that CPCRNet has higher registration accuracy and better robustness compared to existing algorithms.

Chapter 5 - Data Preprocessing and Feature Extraction

An improved lightweight deep neural network with knowledge distillation for local feature extraction and visual localization using images and LiDAR point clouds

For autonomous pipeline robot applications, extracting the features in the pipeline environment such as the 90-degree elbow can greatly reduce the error of the pipeline robot odometer and improve the accuracy of the real-time positioning for autonomous pipeline robot. At present, iterative calculations are used in most of the features extracting methods such as least squares method, but with the huge amount of point cloud data, the computational complexity of these methods is high, and the amount of computation limits the application on embedded robots. For this problem, a network framework For the pipe environment is proposed in this article, which is only for point cloud data input. Based on You Only Look Once v4-tiny(YOLOv4-tiny), a rapid 2D standard detection network framework for images expanding, the discrete 2D point cloud data in the form of bird's eye view is encoded in low-resolution as the input of the net and point of interest (POI) is detected and segmented for the extraction of the elbow features and the accurate estimation of the real-time positioning for the pipeline robot. Our experiments in narrow pipe environment show that compared with the current point cloud feature extraction methods, the proposed method is faster and more accurate.

Nowadays, point clouds are frequently gathered by 3D scanners such as Lidar and Kinect, which produces thousands of point cloud models. Point cloud processing is vital to 3D vision, especially 3D object recognition, positioning, and navigation technology. Addressing uneven data density caused by coordinate frame transformations and the inherent problem of insufficient context connection in point clouds, the DiM-PCNet (Multi-scale and Multi-level Point Clouds Classification Network, and the Di is a prefix to represent double M in Multi-scale and Multi-level) is proposed in this paper. DiM-PCNet is provided for object classification with multi-scale and multi-level features. We encode the point cloud in multi-scale and fully fuse the features with the raw point cloud for keeping the context relationship. In DiM-PCNet, we sample the point clouds from eight parts for multi-scales feature extraction. The multi-scales features are fully fused by multi-level pyramid models. The multi-scale and multi-level strategies are applied in DiM-PCNet, in which the abundant and important features of point clouds are extracted and utilized in the 3D object classification. It is worth noting that the DiM-PCNet feature block can be embedded into the segmentation net, where the accuracy achieved is 87.1%. We conducted experiments on ShapeNet and ModelNet40 and the experimental results show that DiM-PCNet achieves state-of-the-art performance in 3D object classification. The experiment shows competitive performance on robustness and segmentation tasks.

According to the characteristics of the 1-2 coal seam of Shangwan coal mine 4th panel, this research analyzed the feasibility of sublevel caving hydraulic support in the fully-mechanized mining face through estimation method and numerical simulation calculation. This paper also researched the several factors that affect the caving property of top-coal, such as coal thickness, coal hosting depth, top-coal joint fissure, parting condition and caving height. In the meantime, the FLAC3D was used to simulate the caving progress. The results showed that the scheme is feasible. This method cannot only guarantee the stability of coal wall but also crush the coal by the force from the mine pressure hydraulic supports.

LiDAR point cloud-based place recognition (LPR) in unstructured natural environments remains an open challenge with limited existing research. To address the limitations of unstructured environments, such as sparse structural features, uneven point cloud density, and significant viewpoint variations, we present BSPR-Net, a dual-branch point cloud feature extraction approach for point cloud place recognition, which consists of a BEV - projection rotation - invariant convolution branch and a point cloud sparse convolution branch. This design enhances the representation capability of geometric structural features while aggregating rotation-invariant characteristics of point clouds, thereby better addressing the challenge of large viewpoint disparities in reverse-revisited unstructured environments. The proposed network was tested on multiple reverse-revisited sequences of the Wild-Places data set, a benchmark for unstructured natural environment place recognition. It achieved a maximum F1 score of 85.46 %, exceeding other classical methods by more than 4 %. The ablation experiments further confirmed the effectiveness of each module in improving place recognition performance.

Point cloud data, a flexible 3D object representation, is critical for various applications such as autonomous driving, robotics and remote sensing. Despite the recent success of deep neural networks (DNNs) on supervised point cloud analysis tasks, they still rely on tedious manual annotation of point clouds and cannot make predictions for new classes. Unlike few-shot learning for 2D images with the advantages of large-scale datasets and high-quality deep pre-trained models like ResNet, for 3D few-shot learning, obtaining discriminative representations of unseen classes with high intra-class similarity and inter-class difference is very challenging. To address this issue, this work proposes a novel cascade graph neural network for few-shot learning on point clouds, termed as CGNN, in which two cascade GNNs are adopted to extract the intra-object topological information and learn the inter-object relations respectively. To further increase the discriminability of point cloud features, we first design a novel discriminative edge label to model the intra-class similarity and inter-class dissimilarity based on channel-wise feature variance and class consistency. Second, we propose a novel few-shot circle loss which classifies the nodes into two subsets, i.e., support to support pairs and support to query pairs, and optimizes the pair-wise similarity on two subsets independently. Extensive experiments on benchmark CAD and real LiDAR point cloud datasets have demonstrated that CGNN improves accuracy by 5.98% over the state-of-the-art GNN-based few-shot classification methods.

Point cloud completion aims to reconstruct the complete 3D shape from incomplete point clouds, and it is crucial for tasks such as 3D object detection and segmentation. Despite the continuous advances in point cloud analysis techniques, feature extraction methods are still confronted with apparent limitations. The sparse sampling of point clouds, used as inputs in most methods, often results in a certain loss of global structure information. Meanwhile, traditional local feature extraction methods usually struggle to capture the intricate geometric details. To overcome these drawbacks, we introduce PointCFormer, a transformer framework optimized for robust global retention and precise local detail capture in point cloud completion. This framework embraces several key advantages. First, we propose a relation-based local feature extraction method to perceive local delicate geometry characteristics. This approach establishes a fine-grained relationship metric between the target point and its k-nearest neighbors, quantifying each neighboring point's contribution to the target point's local features. Secondly, we introduce a progressive feature extractor that integrates our local feature perception method with self-attention. Starting with a denser sampling of points as input, it iteratively queries long-distance global dependencies and local neighborhood relationships. This extractor maintains enhanced global structure and refined local details, without generating substantial computational overhead. Additionally, we develop a correction module after generating point proxies in the latent space to reintroduce denser information from the input points, enhancing the representation capability of the point proxies. PointCFormer demonstrates state-of-the-art performance on several widely used benchmarks.

A well-known difficulty in processing 3D point cloud is due to its unordered nature. The state-of-the-art 3D computer vision detection methods mainly use voxels to extract point cloud features. However, a drawback of point cloud voxelization is that some features of point cloud surface cannot be fully described. To address this problem, we propose an adaptive sorting approach for voxel-based point cloud. It adaptively ranks point cloud within voxels at first and then goes on to ex-tract voxel features using a multilayer perceptron, called PAS-Net (Points-Adaptive –Sorting-Net). The resulting method is plug-and-play and can be easily deployed in any voxel-based detection model. The proposed model was tested on the KITTI benchmark suit where a 2% − 7% improvement of the detection accuracy is achieved compared with other models using different point feature extraction methods.

The challenges of point cloud registration in intelligent vehicle driving lie in the large scale, complex distribution, high noise, and strong sparsity of lidar point cloud data. This paper proposes an efficient registration algorithm for large-scale outdoor road scenes by selecting the continuous distribution of key area laser point clouds as the registration point cloud. The algorithm extracts feature descriptions of the key point cloud and introduces local geometric features of the point cloud to complete rough and fine registration under constraints of key point clouds and point cloud features. The algorithm is verified through extensive experiments under multiple scenarios, with an average registration time of 0.5831 s and an average accuracy of 0.06996 m, showing significant improvement compared to other algorithms. The algorithm is also validated through real-vehicle experiments, demonstrating strong versatility, reliability, and efficiency. This research has the potential to improve environment perception capabilities of autonomous vehicles by solving the point cloud registration problem in large outdoor scenes.

In the past decade, deep neural networks have achieved significant progress in point cloud learning. However, collecting large-scale precisely-annotated point clouds is extremely laborious and expensive, which hinders the scalability of existing point cloud datasets and poses a bottleneck for efficient exploration of point cloud data in various tasks and applications. Label-efficient learning offers a promising solution by enabling effective deep network training with much-reduced annotation efforts. This paper presents the first comprehensive survey of label-efficient learning of point clouds. We address three critical questions in this emerging research field: i) the importance and urgency of label-efficient learning in point cloud processing, ii) the subfields it encompasses, and iii) the progress achieved in this area. To this end, we propose a taxonomy that organizes label-efficient learning methods based on the data prerequisites provided by different types of labels. We categorize four typical label-efficient learning approaches that significantly reduce point cloud annotation efforts: data augmentation, domain transfer learning, weakly-supervised learning, and pretrained foundation models. For each approach, we outline the problem setup and provide an extensive literature review that showcases relevant progress and challenges. Finally, we share our views on the current research challenges and potential future directions.

Next generation networks pave the way for innovative multimedia services based on immersive 360 video, or extended reality data, such as point clouds (PCs). Herein we address the problem of visual feature extraction on point clouds. To this aim, we introduce a class of filters defined on PCs, called Radial Angular Filters, inspired by the properties of the Human Visual System. The filters are designed to detect radial and angular patterns on the PC, and provide a powerful analysis tool for visual feature extraction on point clouds. For specific design of the filter radial shape, their output is interpreted as the cascade of an angle based modulation of the PC signal followed by the computation of the spectral graph wavelet transform on the modulated signal. The filters open new perspectives in the feature extraction of point clouds acquired in real-life scenarios.