DAPSS: A Novel Network for DOM Assisted Oblique Photography Point Cloud Semantic Segmentation

While most existing advanced large-scale point cloud semantic segmentation methods can accurately identify most large-scale objects, there is still room for improvement in the recognition accuracy of small-scale, low-proportion objects. Compared to point clouds, digital orthophoto maps (DOMs) has a...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhenzhen Song, Mingqiang Guo, Liang Wu, Heng Liu, Ying Huang, Zheng Liu
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11048913/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850069932426919936
author Zhenzhen Song
Mingqiang Guo
Liang Wu
Heng Liu
Ying Huang
Zheng Liu
author_facet Zhenzhen Song
Mingqiang Guo
Liang Wu
Heng Liu
Ying Huang
Zheng Liu
author_sort Zhenzhen Song
collection DOAJ
description While most existing advanced large-scale point cloud semantic segmentation methods can accurately identify most large-scale objects, there is still room for improvement in the recognition accuracy of small-scale, low-proportion objects. Compared to point clouds, digital orthophoto maps (DOMs) has a more structured data format, allowing for better recognition of small-scale surface features. However, in existing projection-based methods, directly mapping images onto point clouds leads to occlusion issues. If image and point cloud features are simply concatenated, it results in feature blurring. Based on this observation, this article proposes a DAPSS network for point cloud semantic segmentation, assisted by prior knowledge constructed from DOM. The pretrained DOM features can provide a broader receptive field as guidance for learning the local context features of point clouds. Vertical occlusion has an issue, making ray-based mapping methods unsuitable. We propose a method that search for the nearest mapped point cloud in spherical space to fill in the occluded point cloud based on the already mapped point cloud. The traditional approach of directly concatenating point cloud features with image features often leads to feature blurring. Therefore, we propose a plug-and-play multimodal feature adaptive fusion module, which can adaptively select and aggregate features from different modalities to reduce redundant information further. In addition, we designed a cascaded multimodal feature deep fusion module to promote deep fusion between different modal features. Experiments on two large datasets demonstrate that DAPSS outperforms current mainstream methods, achieving mean Intersection-over-Union scores of 65.9% and 82.9% on the SansetUrban and SUM-Helsinki datasets, respectively. DAPSS not only effectively addresses the recognition of small-scale surface features, but also resolves the occlusion problems associated with projection-based methods.
format Article
id doaj-art-012bdedfeed34b4baf694767f141ac8b
institution DOAJ
issn 1939-1404
2151-1535
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-012bdedfeed34b4baf694767f141ac8b2025-08-20T02:47:39ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-0118182641828010.1109/JSTARS.2025.358280411048913DAPSS: A Novel Network for DOM Assisted Oblique Photography Point Cloud Semantic SegmentationZhenzhen Song0https://orcid.org/0009-0004-7949-4038Mingqiang Guo1https://orcid.org/0000-0003-4097-4814Liang Wu2https://orcid.org/0000-0003-1877-7179Heng Liu3Ying Huang4Zheng Liu5https://orcid.org/0000-0001-6713-6680School of Geography and Information Engineering, China University of Geosciences, Wuhan, ChinaSchool of Computer Science, China University of Geosciences, Wuhan, ChinaSchool of Computer Science and the School of Geography and Information Engineering, China University of Geosciences, Wuhan, ChinaSchool of Geography and Information Engineering, China University of Geosciences, Wuhan, ChinaWuhan Zondy Cyber Technology Company Ltd., Wuhan, ChinaSchool of Computer Science and the School of Geography and Information Engineering, China University of Geosciences, Wuhan, ChinaWhile most existing advanced large-scale point cloud semantic segmentation methods can accurately identify most large-scale objects, there is still room for improvement in the recognition accuracy of small-scale, low-proportion objects. Compared to point clouds, digital orthophoto maps (DOMs) has a more structured data format, allowing for better recognition of small-scale surface features. However, in existing projection-based methods, directly mapping images onto point clouds leads to occlusion issues. If image and point cloud features are simply concatenated, it results in feature blurring. Based on this observation, this article proposes a DAPSS network for point cloud semantic segmentation, assisted by prior knowledge constructed from DOM. The pretrained DOM features can provide a broader receptive field as guidance for learning the local context features of point clouds. Vertical occlusion has an issue, making ray-based mapping methods unsuitable. We propose a method that search for the nearest mapped point cloud in spherical space to fill in the occluded point cloud based on the already mapped point cloud. The traditional approach of directly concatenating point cloud features with image features often leads to feature blurring. Therefore, we propose a plug-and-play multimodal feature adaptive fusion module, which can adaptively select and aggregate features from different modalities to reduce redundant information further. In addition, we designed a cascaded multimodal feature deep fusion module to promote deep fusion between different modal features. Experiments on two large datasets demonstrate that DAPSS outperforms current mainstream methods, achieving mean Intersection-over-Union scores of 65.9% and 82.9% on the SansetUrban and SUM-Helsinki datasets, respectively. DAPSS not only effectively addresses the recognition of small-scale surface features, but also resolves the occlusion problems associated with projection-based methods.https://ieeexplore.ieee.org/document/11048913/Digital orthophoto map (DOM)feature fusionoblique photogrammetrypoint cloud semantic segmentation
spellingShingle Zhenzhen Song
Mingqiang Guo
Liang Wu
Heng Liu
Ying Huang
Zheng Liu
DAPSS: A Novel Network for DOM Assisted Oblique Photography Point Cloud Semantic Segmentation
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Digital orthophoto map (DOM)
feature fusion
oblique photogrammetry
point cloud semantic segmentation
title DAPSS: A Novel Network for DOM Assisted Oblique Photography Point Cloud Semantic Segmentation
title_full DAPSS: A Novel Network for DOM Assisted Oblique Photography Point Cloud Semantic Segmentation
title_fullStr DAPSS: A Novel Network for DOM Assisted Oblique Photography Point Cloud Semantic Segmentation
title_full_unstemmed DAPSS: A Novel Network for DOM Assisted Oblique Photography Point Cloud Semantic Segmentation
title_short DAPSS: A Novel Network for DOM Assisted Oblique Photography Point Cloud Semantic Segmentation
title_sort dapss a novel network for dom assisted oblique photography point cloud semantic segmentation
topic Digital orthophoto map (DOM)
feature fusion
oblique photogrammetry
point cloud semantic segmentation
url https://ieeexplore.ieee.org/document/11048913/
work_keys_str_mv AT zhenzhensong dapssanovelnetworkfordomassistedobliquephotographypointcloudsemanticsegmentation
AT mingqiangguo dapssanovelnetworkfordomassistedobliquephotographypointcloudsemanticsegmentation
AT liangwu dapssanovelnetworkfordomassistedobliquephotographypointcloudsemanticsegmentation
AT hengliu dapssanovelnetworkfordomassistedobliquephotographypointcloudsemanticsegmentation
AT yinghuang dapssanovelnetworkfordomassistedobliquephotographypointcloudsemanticsegmentation
AT zhengliu dapssanovelnetworkfordomassistedobliquephotographypointcloudsemanticsegmentation