文献类型：期刊文献

英文题名：SPA-Net: An Offset-Free Proposal Network for Individual Tree Segmentation from TLS Data

作者：Zhu, Yunjie[1,2] Wang, Zhihao[1,2] Ye, Qiaolin[1,2] Pang, Lifeng[3] Wang, Qian[1,2] Zheng, Xiaolong[1,2] Hu, Chunhua[1,2]

第一作者：Zhu, Yunjie

通信作者：Hu, CH[1];Hu, CH[2]

机构：[1]Nanjing Forestry Univ, Coll Informat Sci & Technol, Nanjing 210037, Peoples R China;[2]Nanjing Forestry Univ, Coll Artificial Intelligence, Nanjing 210037, Peoples R China;[3]Chinese Acad Forestry, Res Inst Forest Resource Informat Tech, Beijing 100091, Peoples R China

年份：2025

卷号：17

期号：13

外文期刊名：REMOTE SENSING

收录：;EI(收录号：20252918778943);Scopus(收录号：2-s2.0-105010426900);WOS:【SCI-EXPANDED(收录号:WOS:001526289100001)】；

基金：This work was supported by Jiangsu Province Carbon Reaching Peak Carbon Neutral Science and Technology Innovation Special Fund Project (Frontier Foundation), research on the evolution law of carbon sinks in typical wetland ecosystems, the potential for increasing sinks and the regulation mechanism [BK20220016] of Jiangsu Province Science and Technology Department. And the National Key Research and Development Program of China under Grant 2022YFD2201005-03 is supported by the Institute of Resource Information, Chinese Academy of Forestry, and the Jiangsu Province Graduate Research and Practice Innovation Program Project KYCX25_1457 that is supported by Jiangsu Provincial Department of Education.

语种：英文

外文关键词：individual tree segmentation; terrestrial laser scanning; instance segmentation; deep learning

摘要：Individual tree segmentation (ITS) from terrestrial laser scanning (TLS) point clouds is foundational for deriving detailed forest structural parameters, crucial for precision forestry, biomass calculation, and carbon accounting. Conventional ITS algorithms often struggle in complex forest stands due to reliance on heuristic rules and manual feature engineering. Deep learning methodologies proffer more efficacious and automated solutions, but their segmentation accuracy is restricted by imprecise center offset predictions, particularly in intricate forest environments. To address this issue, we proposed a deep learning method, SPA-Net, for achieving tree instance segmentation of forest point clouds. Unlike methods heavily reliant on potentially error-prone global offset vector predictions, SPA-Net employs a novel sampling-shifting-grouping paradigm within its sparse geometric proposal (SGP) module to directly generate initial proposal candidates from raw point data, aiming to reduce dependence on the offset branch. Subsequently, an affinity aggregation (AA) module robustly refines these proposals by assessing inter-proposal relationships and merging fragmented segments, effectively mitigating oversegmentation of large or complex trees; integrating with SGP eliminates the postprocessing step of scoring/NMS. SPA-Net was rigorously validated on two different forest datasets. On both BaiMa and Hong-Tes Lake datasets, the approach demonstrated superior performance compared to several contemporary segmentation approaches evaluated under the same conditions. It achieved 95.8% precision, 96.3% recall, and 92.9% coverage on BaiMa dataset, and achieved 92.6% precision, 94.8% recall, and 88.8% coverage on the Hong-Tes Lake dataset. This study provides a robust tool for individual tree analysis, advancing the accuracy of individual tree segmentation in challenging forest environments.