Urban point cloud classification with automated processing based on deep learning open-source solutions
The recent development of technology in the field of point cloud acquisition in urban environment led experts to consider new needs in terms of big data management in Geomatics. MMS (Mobile Mapping Systems) allows to acquire in real time large number of points in seconds using mobile vehicles. On th...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2024-12-01
|
| Series: | The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences |
| Online Access: | https://isprs-archives.copernicus.org/articles/XLVIII-2-W8-2024/267/2024/isprs-archives-XLVIII-2-W8-2024-267-2024.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The recent development of technology in the field of point cloud acquisition in urban environment led experts to consider new needs in terms of big data management in Geomatics. MMS (Mobile Mapping Systems) allows to acquire in real time large number of points in seconds using mobile vehicles. On the other hand, the development and spread of DL (Deep Learning) CNN (Convolutional Neural Networks) applications for classification of 3D geospatial data allows to start first experimentations in urban environments. Considering this scenario, this work presents a framework for the segmentation and classification of urban point cloud datasets based on the integration of some machine and deep learning tools, namely RANSAC, Euclidean Cluster Extraction, PointNet++ and Support Vector Machines (SVM) algorithms, developed with open-source technologies. The case of study considered in this experimentation regards an urban point cloud dataset obtained by an MMS acquisition along the roads of Sesto Fiorentino (FI). The integration of different algorithms based on Python libraries allowed to obtain fast processing performance, optimizing the results and offering a low-cost and fast solution for experts involved in 3D geospatial data extraction from point cloud MMS acquisitions. |
|---|---|
| ISSN: | 1682-1750 2194-9034 |