Point‐selection and multi‐level‐point‐feature fusion‐based 3D point cloud classification

Zhang, Qieshi; Cheng, Jun; Wang, Shengwen; Xu, Chengjun; Gao, Xiangyang

doi:10.1049/el.2019.2856

Cited by 8 publications

(3 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Point cloud data is different from other data, it is disordered, and the same point may have many different manifestations, which leads to the traditional deep learning that cannot process the point cloud data, and the point cloud data has certain sparsity, and it is difficult to use the deep learning method to process the sparse point cloud when the collection device collects the point cloud data. As shown in the figure, the point cloud consists of four points, f a , f b , f c , and f d , and the point cloud data collected by different devices may have different orders, and the traditional methods will incorrectly identify the point clouds with different orders as different classes [16][17][18][19].…”

Section: Introductionmentioning

confidence: 99%

Point Cloud Key Point Extraction Algorithm Based on Feature Space Value Filtering

Chen

2022

Mobile Information Systems

View full text Add to dashboard Cite

With the rapid development of 3-dimensional (3D) acquisition technology, point clouds have a wide range of application prospects in the fields of computer vision, autonomous driving, and robotics. Point cloud data is widely used in many 3D scenes, and deep learning has become a mainstream research method for classification with the advantages of automatic feature extraction and strong generalization ability. In this paper, a hierarchical key point extraction framework is proposed to solve the problem of modeling the local geometric structure between points. Various point cloud models such as PointNet, PointNet++, and DGCNN are analyzed and their features in local key point are extracted. Based on these analyses, an indexed edge geometric feature spatial value screening neural network (IEGCNN) is proposed. This network extracts features from each point and its neighborhood, calculates the distance between the center point and the points within its neighborhood, and adds the point orientation information to the edge feature spatial value screening network. The relationship between points in the edge network architecture is projected onto a 3D coordinate system and decomposed into three orthogonal bases. The geometric structure between two points is modeled by feature aggregation based on the angle between the edge vector and the base vector and the distance between the center point and the neighboring points. The proposed method has the capability of fast processing of point cloud data by significantly reducing the training and recognition time. The experimental results show that this method achieved high classification accuracy value. This work also provides an idea to solve the problem of real-time target detection network, which has a broad applications prospect in the deployment of movable devices and real-time processing.

show abstract

Section: Introductionmentioning

confidence: 99%

Point Cloud Key Point Extraction Algorithm Based on Feature Space Value Filtering

Chen

2022

Mobile Information Systems

View full text Add to dashboard Cite

show abstract

“…In recent years, the accuracy and robustness of point cloud registration methods based on deep learning have achieved remarkable enhancements with the advancements in hardware and deep learning technology [2,3]. Nevertheless, deep learning methods consume massive computational resources when processing light detection and ranging (LiDAR) point cloud directly, prompting these methods to downsampling the original point cloud prior to input into the network to alleviate this problem [4]. The matching of features is crucial for deep learning-based point cloud registration [5].…”

mentioning

confidence: 99%

MDU‐sampling: Multi‐domain uniform sampling method for large‐scale outdoor LiDAR point cloud registration

Ou,

Zheng,

Zheng

2024

Electronics Letters

View full text Add to dashboard Cite

Sampling is a crucial concern for outdoor light detection and ranging (LiDAR) point cloud registration due to the large amounts of point cloud. Numerous algorithms have been devised to tackle this issue by selecting key points. However, these approaches often necessitate extensive computations, giving rise to challenges related to computational time and complexity. This letter proposes a multi‐domain uniform sampling method (MDU‐sampling) for large‐scale outdoor LiDAR point cloud registration. The feature extraction based on deep learning aggregates information from the neighbourhood, so there is redundancy between adjacent features. The sampling method in this paper is carried out in the spatial and feature domains. First, uniform sampling is executed in the spatial domain, maintaining local point cloud uniformity. This is believed to preserve more potential point correspondences and is beneficial for subsequent neighbourhood information aggregation and feature sampling. Subsequently, a secondary sampling in the feature domain is performed to reduce redundancy among the features of neighbouring points. Notably, only points on the same ring in LiDAR data are considered as neighbouring points, eliminating the need for additional neighbouring point search and thereby speeding up processing rates. Experimental results demonstrate that the approach enhances accuracy and robustness compared with benchmarks.

show abstract

“…As a result, the motion-based methods can ensure the robustness of the tracker against interference and changes in appearance. In contrast to existing methods [13][14][15], in this letter, we enhance feature learning by introducing the point position embedding module followed by a self-attention coding module. These modules allow us to capture the spatial relationship between points within point clouds more effectively.…”

mentioning

confidence: 99%

PPE: Point position embedding for single object tracking in point clouds

Wang

et al. 2023

Electronics Letters

View full text Add to dashboard Cite

Existing 3D single object tracking methods primarily extract features from the global coordinates of point clouds, overlooking the potential exploitation of their positional information. However, due to the unordered, sparse, and irregular nature of point clouds, effectively exploring their positional information presents a significant challenge. In this letter, the network is explicitly reformulated by introducing a point position embedding module in conjunction with a self‐attention coding module, replacing the use of global coordinate inputs. The proposed reformulation is further integrated into a top‐notch model M2‐Track, called Point Position Embedding (PPE) in this letter. Comprehensive empirical analysis are performed on the KITTI and NuScenes datasets. Experimental results show that the PPE surpasses M2‐Track by a large margin in overall performance. Especially for the challenging NuScenes dataset, the method attains the highest precision and success in all classes compared to state‐of‐the‐art methods. The code is available at https://github.com/GZHU‐DVL/PPE.

show abstract

Point‐selection and multi‐level‐point‐feature fusion‐based 3D point cloud classification

Cited by 8 publications

References 11 publications

Point Cloud Key Point Extraction Algorithm Based on Feature Space Value Filtering

Point Cloud Key Point Extraction Algorithm Based on Feature Space Value Filtering

MDU‐sampling: Multi‐domain uniform sampling method for large‐scale outdoor LiDAR point cloud registration

PPE: Point position embedding for single object tracking in point clouds

Contact Info

Product

Resources

About