Evaluation of Global-Scale and Local-Scale Optimized Segmentation Algorithms in GEOBIA With SAM on Land Use and Land Cover

Segmentation is crucial in geographic object-based image analysis for accurate land use and land cover mapping. However, obtaining outstanding segmentation results in all scenarios proves challenging with a single algorithm. This study investigates seven segmentation algorithms: mean shift (MF), O S...

Full description

Saved in:
Bibliographic Details
Main Authors: Tao He, Jianyu Chen, Linchong Kang, Qiankun Zhu
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10460074/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850041570851553280
author Tao He
Jianyu Chen
Linchong Kang
Qiankun Zhu
author_facet Tao He
Jianyu Chen
Linchong Kang
Qiankun Zhu
author_sort Tao He
collection DOAJ
description Segmentation is crucial in geographic object-based image analysis for accurate land use and land cover mapping. However, obtaining outstanding segmentation results in all scenarios proves challenging with a single algorithm. This study investigates seven segmentation algorithms: mean shift (MF), O Sistema de Processamento de informa&#x00E7;&#x00F5;es georreferenciadas (the geographic information and image processing system) (SPRING), Estimation of scale parameter 2 (ESP2) (three global-scale algorithms), image object detection approach (IODA), SA, edge-guided image object detection approach (EIODA) (three local-scale optimization algorithms), and segment anything model (SAM) (deep learning). In the custom dataset and semantic segmentation datasets, we apply visual interpretation, unsupervised, and supervised evaluation methods with 15 test images, using a total of 17 evaluation indices to assess the segmentation results. Based on the evaluation results, the effectiveness and adaptability of the algorithms in scene segmentation are comprehensively analyzed. The results report that global-scale segmentation approaches encounter difficulties in distinguishing meaningful objects in complicated scenarios. Both MF and SPRING methods are prone to over-segmentation. In many cases, ESP2 tends to generate homogeneous segments (low weighted variance), whereas EIODA tends to produce heterogeneous adjacent segments (low Moran&#x0027;s I). ED3 and segmentation evaluation index demonstrate that scale parameter (SA) and IODA can to some extent identify geo-objects, with SA being more effective and performing exceptionally well in building extraction. The EIODA performs well in areas with clear boundaries, like aquaculture ponds and water-land transitions. SAM accurately detects objects of various sizes, displaying rich semantic content and high consistency with reference polygons. The average intersection over union reaches 71.10&#x0025; and <italic>F</italic> measure attains 0.77 under normal conditions.
format Article
id doaj-art-23e651abb8854903abce357bd487c2a9
institution DOAJ
issn 1939-1404
2151-1535
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-23e651abb8854903abce357bd487c2a92025-08-20T02:55:45ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352024-01-01176721673810.1109/JSTARS.2024.337338510460074Evaluation of Global-Scale and Local-Scale Optimized Segmentation Algorithms in GEOBIA With SAM on Land Use and Land CoverTao He0https://orcid.org/0009-0008-6602-2305Jianyu Chen1https://orcid.org/0000-0002-2354-0240Linchong Kang2https://orcid.org/0009-0003-8567-3959Qiankun Zhu3https://orcid.org/0009-0008-5069-8052Ocean College, Zhejiang University, Zhoushan, ChinaState Key Laboratory of Satellite Ocean Environment Dynamics, Second Institute of Oceanography, Ministry of Natural Resources, Hangzhou, ChinaNational Marine Data and Information Service, Tianjin, ChinaState Key Laboratory of Satellite Ocean Environment Dynamics, Second Institute of Oceanography, Ministry of Natural Resources, Hangzhou, ChinaSegmentation is crucial in geographic object-based image analysis for accurate land use and land cover mapping. However, obtaining outstanding segmentation results in all scenarios proves challenging with a single algorithm. This study investigates seven segmentation algorithms: mean shift (MF), O Sistema de Processamento de informa&#x00E7;&#x00F5;es georreferenciadas (the geographic information and image processing system) (SPRING), Estimation of scale parameter 2 (ESP2) (three global-scale algorithms), image object detection approach (IODA), SA, edge-guided image object detection approach (EIODA) (three local-scale optimization algorithms), and segment anything model (SAM) (deep learning). In the custom dataset and semantic segmentation datasets, we apply visual interpretation, unsupervised, and supervised evaluation methods with 15 test images, using a total of 17 evaluation indices to assess the segmentation results. Based on the evaluation results, the effectiveness and adaptability of the algorithms in scene segmentation are comprehensively analyzed. The results report that global-scale segmentation approaches encounter difficulties in distinguishing meaningful objects in complicated scenarios. Both MF and SPRING methods are prone to over-segmentation. In many cases, ESP2 tends to generate homogeneous segments (low weighted variance), whereas EIODA tends to produce heterogeneous adjacent segments (low Moran&#x0027;s I). ED3 and segmentation evaluation index demonstrate that scale parameter (SA) and IODA can to some extent identify geo-objects, with SA being more effective and performing exceptionally well in building extraction. The EIODA performs well in areas with clear boundaries, like aquaculture ponds and water-land transitions. SAM accurately detects objects of various sizes, displaying rich semantic content and high consistency with reference polygons. The average intersection over union reaches 71.10&#x0025; and <italic>F</italic> measure attains 0.77 under normal conditions.https://ieeexplore.ieee.org/document/10460074/Geographic object-based image analysis (GEOBIA)scale optimizationsegment anything model (SAM)segment evaluation
spellingShingle Tao He
Jianyu Chen
Linchong Kang
Qiankun Zhu
Evaluation of Global-Scale and Local-Scale Optimized Segmentation Algorithms in GEOBIA With SAM on Land Use and Land Cover
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Geographic object-based image analysis (GEOBIA)
scale optimization
segment anything model (SAM)
segment evaluation
title Evaluation of Global-Scale and Local-Scale Optimized Segmentation Algorithms in GEOBIA With SAM on Land Use and Land Cover
title_full Evaluation of Global-Scale and Local-Scale Optimized Segmentation Algorithms in GEOBIA With SAM on Land Use and Land Cover
title_fullStr Evaluation of Global-Scale and Local-Scale Optimized Segmentation Algorithms in GEOBIA With SAM on Land Use and Land Cover
title_full_unstemmed Evaluation of Global-Scale and Local-Scale Optimized Segmentation Algorithms in GEOBIA With SAM on Land Use and Land Cover
title_short Evaluation of Global-Scale and Local-Scale Optimized Segmentation Algorithms in GEOBIA With SAM on Land Use and Land Cover
title_sort evaluation of global scale and local scale optimized segmentation algorithms in geobia with sam on land use and land cover
topic Geographic object-based image analysis (GEOBIA)
scale optimization
segment anything model (SAM)
segment evaluation
url https://ieeexplore.ieee.org/document/10460074/
work_keys_str_mv AT taohe evaluationofglobalscaleandlocalscaleoptimizedsegmentationalgorithmsingeobiawithsamonlanduseandlandcover
AT jianyuchen evaluationofglobalscaleandlocalscaleoptimizedsegmentationalgorithmsingeobiawithsamonlanduseandlandcover
AT linchongkang evaluationofglobalscaleandlocalscaleoptimizedsegmentationalgorithmsingeobiawithsamonlanduseandlandcover
AT qiankunzhu evaluationofglobalscaleandlocalscaleoptimizedsegmentationalgorithmsingeobiawithsamonlanduseandlandcover