CEDAnet: Individual Tree Segmentation in Dense Orchard via Context Enhancement and Density Prior
Individual tree segmentation (ITS) is a pivotal technique in orchard research, estimating tree counts and delineating crown contours. This method provides foundational data for assessing orchard health, nutritional composition, and predicting yield. Unmanned aerial vehicles (UAVs) have become an ess...
| Main Authors: | , , , , |
|---|---|
| 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/10475902/ |
| _version_ | 1797222050292563968 |
|---|---|
| author | Fangjie Zhu Zhenhao Chen Haoyang Li Qian Shi Xiaoping Liu |
| author_facet | Fangjie Zhu Zhenhao Chen Haoyang Li Qian Shi Xiaoping Liu |
| author_sort | Fangjie Zhu |
| collection | DOAJ |
| description | Individual tree segmentation (ITS) is a pivotal technique in orchard research, estimating tree counts and delineating crown contours. This method provides foundational data for assessing orchard health, nutritional composition, and predicting yield. Unmanned aerial vehicles (UAVs) have become an essential data source for (ITS) due to their capability to capture ultra-fine details. However, current deep-learning-based ITS methods struggle to accurately handle densely overlapping fruit tree distributions with similar characteristics in UAVs images, primarily due to the intricate nature of spatial arrangements in such scenarios. In this article, we propose CEDAnet, a context enhancement, and density adjustment network, to address the challenge of dense fruit trees segmentation. Specifically, a transformer-based contextual aggregation module is designed to distinguish different instances and refine the boundary of the instances. We have proposed a density-guided nonmaximum suppression method to adaptively generate sufficient candidate bounding boxes, aiming to retain more potential instances in dense trees. To evaluate the effectiveness and robustness of our proposal, we curated two ITS datasets constructed with imagery captured by UAVs, namely instance segmentation in Conghua images dataset (iSCHID) and instance segmentation in Maoming images dataset (iSMMID) based on their respective spatial characteristics. Experimental results on both two datasets demonstrated that CEDAnet yields competitive results in ITS tasks, with the bounding box AP of 0.498, segmentation AP of 0.493 in iSCHID, and the bounding box AP of 0.706, segmentation AP of 0.703 in iSMMID. |
| first_indexed | 2024-04-24T13:15:10Z |
| format | Article |
| id | doaj.art-8dafc3206be14568ab5043d2b4680c61 |
| institution | Directory Open Access Journal |
| issn | 2151-1535 |
| language | English |
| last_indexed | 2024-04-24T13:15:10Z |
| 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-8dafc3206be14568ab5043d2b4680c612024-04-04T23:00:15ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352024-01-01177040705110.1109/JSTARS.2024.337816710475902CEDAnet: Individual Tree Segmentation in Dense Orchard via Context Enhancement and Density PriorFangjie Zhu0https://orcid.org/0009-0005-5756-0708Zhenhao Chen1https://orcid.org/0009-0004-0560-728XHaoyang Li2https://orcid.org/0000-0001-8725-342XQian Shi3https://orcid.org/0000-0002-1276-0352Xiaoping Liu4https://orcid.org/0000-0003-4242-5392Guangdong Provincial Key Laboratory for Urbanization and GeoSimulation, School of Geography and Planning, Sun Yat-sen University, Guangzhou, ChinaGuangdong Provincial Key Laboratory for Urbanization and GeoSimulation, School of Geography and Planning, Sun Yat-sen University, Guangzhou, ChinaGuangdong Provincial Key Laboratory for Urbanization and GeoSimulation, School of Geography and Planning, Sun Yat-sen University, Guangzhou, ChinaGuangdong Provincial Key Laboratory for Urbanization and GeoSimulation, School of Geography and Planning, Sun Yat-sen University, Guangzhou, ChinaGuangdong Provincial Key Laboratory for Urbanization and GeoSimulation, School of Geography and Planning, Sun Yat-sen University, Guangzhou, ChinaIndividual tree segmentation (ITS) is a pivotal technique in orchard research, estimating tree counts and delineating crown contours. This method provides foundational data for assessing orchard health, nutritional composition, and predicting yield. Unmanned aerial vehicles (UAVs) have become an essential data source for (ITS) due to their capability to capture ultra-fine details. However, current deep-learning-based ITS methods struggle to accurately handle densely overlapping fruit tree distributions with similar characteristics in UAVs images, primarily due to the intricate nature of spatial arrangements in such scenarios. In this article, we propose CEDAnet, a context enhancement, and density adjustment network, to address the challenge of dense fruit trees segmentation. Specifically, a transformer-based contextual aggregation module is designed to distinguish different instances and refine the boundary of the instances. We have proposed a density-guided nonmaximum suppression method to adaptively generate sufficient candidate bounding boxes, aiming to retain more potential instances in dense trees. To evaluate the effectiveness and robustness of our proposal, we curated two ITS datasets constructed with imagery captured by UAVs, namely instance segmentation in Conghua images dataset (iSCHID) and instance segmentation in Maoming images dataset (iSMMID) based on their respective spatial characteristics. Experimental results on both two datasets demonstrated that CEDAnet yields competitive results in ITS tasks, with the bounding box AP of 0.498, segmentation AP of 0.493 in iSCHID, and the bounding box AP of 0.706, segmentation AP of 0.703 in iSMMID.https://ieeexplore.ieee.org/document/10475902/Benchmark datasetdeep learning (DL)individual tree segmentation (ITS)instance segmentationunmanned aerial vehicle (UAV) |
| spellingShingle | Fangjie Zhu Zhenhao Chen Haoyang Li Qian Shi Xiaoping Liu CEDAnet: Individual Tree Segmentation in Dense Orchard via Context Enhancement and Density Prior IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Benchmark dataset deep learning (DL) individual tree segmentation (ITS) instance segmentation unmanned aerial vehicle (UAV) |
| title | CEDAnet: Individual Tree Segmentation in Dense Orchard via Context Enhancement and Density Prior |
| title_full | CEDAnet: Individual Tree Segmentation in Dense Orchard via Context Enhancement and Density Prior |
| title_fullStr | CEDAnet: Individual Tree Segmentation in Dense Orchard via Context Enhancement and Density Prior |
| title_full_unstemmed | CEDAnet: Individual Tree Segmentation in Dense Orchard via Context Enhancement and Density Prior |
| title_short | CEDAnet: Individual Tree Segmentation in Dense Orchard via Context Enhancement and Density Prior |
| title_sort | cedanet individual tree segmentation in dense orchard via context enhancement and density prior |
| topic | Benchmark dataset deep learning (DL) individual tree segmentation (ITS) instance segmentation unmanned aerial vehicle (UAV) |
| url | https://ieeexplore.ieee.org/document/10475902/ |
| work_keys_str_mv | AT fangjiezhu cedanetindividualtreesegmentationindenseorchardviacontextenhancementanddensityprior AT zhenhaochen cedanetindividualtreesegmentationindenseorchardviacontextenhancementanddensityprior AT haoyangli cedanetindividualtreesegmentationindenseorchardviacontextenhancementanddensityprior AT qianshi cedanetindividualtreesegmentationindenseorchardviacontextenhancementanddensityprior AT xiaopingliu cedanetindividualtreesegmentationindenseorchardviacontextenhancementanddensityprior |