Finding the Optimal Convolutional Kernel Size for Semantic Segmentation of Pole-like Objects in Lidar Point Clouds

Pole-like objects (PLOs) are important street assets in urban environments, yet current deep learning methods often underperform in their segmentation compared to other objects. The main challenge is determining the right kernel size to effectively understand the unique structure of PLOs with an app...

Full description

Saved in:
Bibliographic Details
Main Authors: Z. Zhang, D. Shojaei, K. Khoshelham
Format: Article
Language:English
Published: Copernicus Publications 2025-08-01
Series:The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Online Access:https://isprs-archives.copernicus.org/articles/XLVIII-G-2025/1747/2025/isprs-archives-XLVIII-G-2025-1747-2025.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849245192028160000
author Z. Zhang
D. Shojaei
K. Khoshelham
author_facet Z. Zhang
D. Shojaei
K. Khoshelham
author_sort Z. Zhang
collection DOAJ
description Pole-like objects (PLOs) are important street assets in urban environments, yet current deep learning methods often underperform in their segmentation compared to other objects. The main challenge is determining the right kernel size to effectively understand the unique structure of PLOs with an appropriate receptive field. In this study, we improve the segmentation performance of PLOs by optimizing the kernel size in a KPConv-based network. Our experiments show that kernel size of 9 yields an Intersection over Union (IoU) of 95.02% on the Parkville-3D dataset. We also develop a post-processing approach that transforms semantic segmentation outputs into panoptic segmentation results, enabling accurate detection of individual PLO instances. Furthermore, qualitative tests on an independent, unlabelled point cloud dataset from a different urban area demonstrate that our method consistently achieves accurate segmentation.
format Article
id doaj-art-5be3dad5ba254f8580ef99daaeaef8a5
institution Kabale University
issn 1682-1750
2194-9034
language English
publishDate 2025-08-01
publisher Copernicus Publications
record_format Article
series The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
spelling doaj-art-5be3dad5ba254f8580ef99daaeaef8a52025-08-20T03:58:54ZengCopernicus PublicationsThe International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences1682-17502194-90342025-08-01XLVIII-G-20251747175210.5194/isprs-archives-XLVIII-G-2025-1747-2025Finding the Optimal Convolutional Kernel Size for Semantic Segmentation of Pole-like Objects in Lidar Point CloudsZ. Zhang0D. Shojaei1K. Khoshelham2Centre for Spatial Data Infrastructures and Land Administration, Department of Infrastructure Engineering, The University of Melbourne, Melbourne, Victoria, AustraliaCentre for Spatial Data Infrastructures and Land Administration, Department of Infrastructure Engineering, The University of Melbourne, Melbourne, Victoria, AustraliaDepartment of Infrastructure Engineering, The University of Melbourne, Melbourne, Victoria, AustraliaPole-like objects (PLOs) are important street assets in urban environments, yet current deep learning methods often underperform in their segmentation compared to other objects. The main challenge is determining the right kernel size to effectively understand the unique structure of PLOs with an appropriate receptive field. In this study, we improve the segmentation performance of PLOs by optimizing the kernel size in a KPConv-based network. Our experiments show that kernel size of 9 yields an Intersection over Union (IoU) of 95.02% on the Parkville-3D dataset. We also develop a post-processing approach that transforms semantic segmentation outputs into panoptic segmentation results, enabling accurate detection of individual PLO instances. Furthermore, qualitative tests on an independent, unlabelled point cloud dataset from a different urban area demonstrate that our method consistently achieves accurate segmentation.https://isprs-archives.copernicus.org/articles/XLVIII-G-2025/1747/2025/isprs-archives-XLVIII-G-2025-1747-2025.pdf
spellingShingle Z. Zhang
D. Shojaei
K. Khoshelham
Finding the Optimal Convolutional Kernel Size for Semantic Segmentation of Pole-like Objects in Lidar Point Clouds
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
title Finding the Optimal Convolutional Kernel Size for Semantic Segmentation of Pole-like Objects in Lidar Point Clouds
title_full Finding the Optimal Convolutional Kernel Size for Semantic Segmentation of Pole-like Objects in Lidar Point Clouds
title_fullStr Finding the Optimal Convolutional Kernel Size for Semantic Segmentation of Pole-like Objects in Lidar Point Clouds
title_full_unstemmed Finding the Optimal Convolutional Kernel Size for Semantic Segmentation of Pole-like Objects in Lidar Point Clouds
title_short Finding the Optimal Convolutional Kernel Size for Semantic Segmentation of Pole-like Objects in Lidar Point Clouds
title_sort finding the optimal convolutional kernel size for semantic segmentation of pole like objects in lidar point clouds
url https://isprs-archives.copernicus.org/articles/XLVIII-G-2025/1747/2025/isprs-archives-XLVIII-G-2025-1747-2025.pdf
work_keys_str_mv AT zzhang findingtheoptimalconvolutionalkernelsizeforsemanticsegmentationofpolelikeobjectsinlidarpointclouds
AT dshojaei findingtheoptimalconvolutionalkernelsizeforsemanticsegmentationofpolelikeobjectsinlidarpointclouds
AT kkhoshelham findingtheoptimalconvolutionalkernelsizeforsemanticsegmentationofpolelikeobjectsinlidarpointclouds