Classification of LiDAR Point Cloud based on Multiscale Features and PointNet

Zhao, Zhijie; Cheng, Yinglei; Xiaosong, Shi; Qin, Xianxiang; Li, Sun

doi:10.1109/ipta.2018.8608120

Cited by 5 publications

(3 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Yet, the network does not learn the connection among local points; hence, the model cannot capture information on local features. Zhao et al [24] proposed a deep neural network that combines multiscale features with PointNet, adopting multiscale approaches to extract the neighborhood features of points and combining them with global features extracted by PointNet to classify LiDAR point clouds. The network has a good classification effect; however, it extracts local features with poor efficiency.…”

Section: Point Cloud-basedmentioning

confidence: 99%

Fusion of a Static and Dynamic Convolutional Neural Network for Multiview 3D Point Cloud Classification

et al. 2022

View full text Add to dashboard Cite

Three-dimensional (3D) point cloud classification methods based on deep learning have good classification performance; however, they adapt poorly to diverse datasets and their classification accuracy must be improved. Therefore, FSDCNet, a neural network model based on the fusion of static and dynamic convolution, is proposed and applied for multiview 3D point cloud classification in this paper. FSDCNet devises a view selection method with fixed and random viewpoints, which effectively avoids the overfitting caused by the traditional fixed viewpoint. A local feature extraction operator of dynamic and static convolution adaptive weight fusion was designed to improve the model’s adaptability to different types of datasets. To address the problems of large parameters and high computational complexity associated with the current methods of dynamic convolution, a lightweight and adaptive dynamic convolution operator was developed. In addition, FSDCNet builds a global attention pooling, integrating the most crucial information on different view features to the greatest extent. Due to these characteristics, FSDCNet is more adaptable, can extract more fine-grained detailed information, and can improve the classification accuracy of point cloud data. The proposed method was applied to the ModelNet40 and Sydney Urban Objects datasets. In these experiments, FSDCNet outperformed its counterparts, achieving state-of-the-art point cloud classification accuracy. For the ModelNet40 dataset, the overall accuracy (OA) and average accuracy (AA) of FSDCNet in a single view reached 93.8% and 91.2%, respectively, which were superior to those values for many other methods using 6 and 12 views. FSDCNet obtained the best results for 6 and 12 views, achieving 94.6%, 93.3%, 95.3%, and 93.6% in OA and AA metrics, respectively. For the Sydney Urban Objects dataset, FSDCNet achieved an OA and F1 score of 81.2% and 80.1% in a single view, respectively, which were higher than most of the compared methods. In 6 and 12 views, FSDCNet reached an OA of 85.3% and 83.6% and an F1 score of 85.5% and 83.7%, respectively.

show abstract

Section: Point Cloud-basedmentioning

confidence: 99%

Fusion of a Static and Dynamic Convolutional Neural Network for Multiview 3D Point Cloud Classification

et al. 2022

View full text Add to dashboard Cite

show abstract

“…In fact, several ML techniques have been utilized to classify trees, electric poles, buildings, traffic signs, and road surfaces. The primary ones include RF for road edges and traffic signs on road surfaces [23], SVM for objects in urban areas [24], and CNN for objects from point clouds [25].…”

Section: Introductionmentioning

confidence: 99%

Classification of 3D Point Cloud Data from Mobile Mapping System for Detecting Road Surfaces and Potholes using Convolution Neural Networks

2022

IJG

View full text Add to dashboard Cite

Normally, road damages can be automatically detected using image and video data from ground survey vehicle system combined with the detection algorithms. However, there are limitations of scales and map coordinates when using the image and video data to detect potholes. It has been challenging to detect and determine the sizes and locations of potholes. This research utilized a mobile mapping system, MMS, to collect data of roads and environment and classify potholes, roads and other objects. A convolution neural network (CNN) was used to directly identify 3D point clouds using the XYZ method in comparison with the proposed XYZ-RGB method. The XYZ classification demonstrated an overall accuracy of 96.77%, with the intersection over union (IoU) of potholes, roads, and other objects of 59.50%, 94.22%, and 94.06%, respectively. The proposed XYZ-RGB classification indicated an overall accuracy of 97.50%, with the IoU of potholes, roads, and other objects of 66.66%, 95.43%, and 95.42%, respectively. Both datasets were statistically compared at the 95% confidence level, and the results revealed that both classifications produced significantly different results.

show abstract

“…The model knowledge increases in each iteration. Similar work of classifying unstructured point clouds with MLPs can be found in (Soilan et al, 2019), (Winiwarter et al, 2019), (Zhao et al, 2018), (Lian et al, 2019), (Hu et al, 2020), (Zhao et al, 2021) et al…”

Section: Deep Learning-based Classificationmentioning

confidence: 68%

Photogrammetric point clouds: quality assessment, filtering, and change detection

Zhang

View full text Add to dashboard Cite

show abstract

Classification of LiDAR Point Cloud based on Multiscale Features and PointNet

Cited by 5 publications

References 9 publications

Fusion of a Static and Dynamic Convolutional Neural Network for Multiview 3D Point Cloud Classification

Fusion of a Static and Dynamic Convolutional Neural Network for Multiview 3D Point Cloud Classification

Classification of 3D Point Cloud Data from Mobile Mapping System for Detecting Road Surfaces and Potholes using Convolution Neural Networks

Photogrammetric point clouds: quality assessment, filtering, and change detection

Contact Info

Product

Resources

About