SEMANTIC POINT CLOUD SEGMENTATION IN URBAN ENVIRONMENTS WITH 1D CONVOLUTIONAL NEURAL NETWORKS
Convolutional Neural Networks (CNNs) have been widely recognized for their efficacy in image analysis tasks. This paper investigates the application of the 1D-CNN variant CNNs for the semantic segmentation of urban point clouds obtained through Mobile Laser Scanning. Ten well-known local geometric f...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2024-03-01
|
| Series: | The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences |
| Online Access: | https://isprs-archives.copernicus.org/articles/XLVIII-4-W9-2024/205/2024/isprs-archives-XLVIII-4-W9-2024-205-2024.pdf |
| _version_ | 1797268074516185088 |
|---|---|
| author | S. M. González-Collazo N. Canedo-González E. González J. Balado |
| author_facet | S. M. González-Collazo N. Canedo-González E. González J. Balado |
| author_sort | S. M. González-Collazo |
| collection | DOAJ |
| description | Convolutional Neural Networks (CNNs) have been widely recognized for their efficacy in image analysis tasks. This paper investigates the application of the 1D-CNN variant CNNs for the semantic segmentation of urban point clouds obtained through Mobile Laser Scanning. Ten well-known local geometric features of point clouds were used as input for the 1D CNN. Through an empirical analysis on the Santiago Urban Dataset, the 1D CNN was optimized in terms of numbers of convolution layers, neurons, pooling layers, dropout layers, dense layers, training epochs, and batch size. The performance of the proposed 1D CNN was compared with Support Vector Machine (SVM), Random Forest (RF), and PointNet++. Despite demonstrating a F1-score weighted at 70.3%, outperforming SVM but slightly lagging RF (71.6%) and significantly trailing PointNet++ (90.3%), the proposed 1D-CNN showcases a cost-effective potential for the segmentation of <em>road</em> and <em>building</em> classes. The relative computational requirements of the models were also discussed, highlighting the practical advantages and limitations of each approach. |
| first_indexed | 2024-04-25T01:26:42Z |
| format | Article |
| id | doaj.art-49e649a6dcd640efa5c5cee13d25ab82 |
| institution | Directory Open Access Journal |
| issn | 1682-1750 2194-9034 |
| language | English |
| last_indexed | 2024-04-25T01:26:42Z |
| publishDate | 2024-03-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences |
| spelling | doaj.art-49e649a6dcd640efa5c5cee13d25ab822024-03-08T19:20:14ZengCopernicus PublicationsThe International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences1682-17502194-90342024-03-01XLVIII-4-W9-202420521110.5194/isprs-archives-XLVIII-4-W9-2024-205-2024SEMANTIC POINT CLOUD SEGMENTATION IN URBAN ENVIRONMENTS WITH 1D CONVOLUTIONAL NEURAL NETWORKSS. M. González-Collazo0N. Canedo-González1E. González2J. Balado3GeoTECH, CINTECX, Universidade de Vigo, 36310 Vigo, SpainGeoTECH, CINTECX, Universidade de Vigo, 36310 Vigo, SpainGeoTECH, CINTECX, Universidade de Vigo, 36310 Vigo, SpainGeoTECH, CINTECX, Universidade de Vigo, 36310 Vigo, SpainConvolutional Neural Networks (CNNs) have been widely recognized for their efficacy in image analysis tasks. This paper investigates the application of the 1D-CNN variant CNNs for the semantic segmentation of urban point clouds obtained through Mobile Laser Scanning. Ten well-known local geometric features of point clouds were used as input for the 1D CNN. Through an empirical analysis on the Santiago Urban Dataset, the 1D CNN was optimized in terms of numbers of convolution layers, neurons, pooling layers, dropout layers, dense layers, training epochs, and batch size. The performance of the proposed 1D CNN was compared with Support Vector Machine (SVM), Random Forest (RF), and PointNet++. Despite demonstrating a F1-score weighted at 70.3%, outperforming SVM but slightly lagging RF (71.6%) and significantly trailing PointNet++ (90.3%), the proposed 1D-CNN showcases a cost-effective potential for the segmentation of <em>road</em> and <em>building</em> classes. The relative computational requirements of the models were also discussed, highlighting the practical advantages and limitations of each approach.https://isprs-archives.copernicus.org/articles/XLVIII-4-W9-2024/205/2024/isprs-archives-XLVIII-4-W9-2024-205-2024.pdf |
| spellingShingle | S. M. González-Collazo N. Canedo-González E. González J. Balado SEMANTIC POINT CLOUD SEGMENTATION IN URBAN ENVIRONMENTS WITH 1D CONVOLUTIONAL NEURAL NETWORKS The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences |
| title | SEMANTIC POINT CLOUD SEGMENTATION IN URBAN ENVIRONMENTS WITH 1D CONVOLUTIONAL NEURAL NETWORKS |
| title_full | SEMANTIC POINT CLOUD SEGMENTATION IN URBAN ENVIRONMENTS WITH 1D CONVOLUTIONAL NEURAL NETWORKS |
| title_fullStr | SEMANTIC POINT CLOUD SEGMENTATION IN URBAN ENVIRONMENTS WITH 1D CONVOLUTIONAL NEURAL NETWORKS |
| title_full_unstemmed | SEMANTIC POINT CLOUD SEGMENTATION IN URBAN ENVIRONMENTS WITH 1D CONVOLUTIONAL NEURAL NETWORKS |
| title_short | SEMANTIC POINT CLOUD SEGMENTATION IN URBAN ENVIRONMENTS WITH 1D CONVOLUTIONAL NEURAL NETWORKS |
| title_sort | semantic point cloud segmentation in urban environments with 1d convolutional neural networks |
| url | https://isprs-archives.copernicus.org/articles/XLVIII-4-W9-2024/205/2024/isprs-archives-XLVIII-4-W9-2024-205-2024.pdf |
| work_keys_str_mv | AT smgonzalezcollazo semanticpointcloudsegmentationinurbanenvironmentswith1dconvolutionalneuralnetworks AT ncanedogonzalez semanticpointcloudsegmentationinurbanenvironmentswith1dconvolutionalneuralnetworks AT egonzalez semanticpointcloudsegmentationinurbanenvironmentswith1dconvolutionalneuralnetworks AT jbalado semanticpointcloudsegmentationinurbanenvironmentswith1dconvolutionalneuralnetworks |