面向濱海生態(tài)監(jiān)管的多尺度目標(biāo)語義分割研究

Multi-scale Objectives Semantic Segmentation for Coastal Ecological Supervision

Environment and Resource / 2022,4(2)：48-61 / 2022-07-13 458 809

查看全文 下載文檔

• 作者： 陳巖1      楊曉彤2      奚硯濤3      徐立祥1      李新路1     
• 單位：

  1．合肥學(xué)院人工智能與大數(shù)據(jù)學(xué)院，合肥；
  2．國家海洋信息中心海洋測繪地理信息部，天津；
  3．中國礦業(yè)大學(xué)資源與地球科學(xué)學(xué)院，徐州

• 關(guān)鍵詞： 濱海生態(tài)監(jiān)管；人工智能；紅樹林；海水養(yǎng)殖；基準(zhǔn)數(shù)據(jù)集；多尺度特征融合； 語義分割
• Coastal ecological supervision; Artificial intelligence; Mangroves; Mariculture; Benchmark dataset; Multi-scale feature fusion; Semantic segmentation
• 摘要： 針對缺少濱海生態(tài)場景深度學(xué)習(xí)數(shù)據(jù)集，面向遙感影像分類的多尺度目標(biāo)語義分割精度不高等問題，研究以紅樹林、浮筏養(yǎng)殖和圍塘養(yǎng)殖三類濱海典型生態(tài)監(jiān)管多尺度目標(biāo)為研究對象，構(gòu)建了面向濱海生態(tài)監(jiān)管的多目標(biāo)語義分割數(shù)據(jù)集，通過集成批歸一化和空間置棄算法，改進(jìn)UNet 特征融合策略，提出了一種多尺 度深度卷積語義分割模型。模型在測試集上總體精度92%，Kappa 系數(shù)0.87，平均交并比82%。實驗結(jié)果表明批歸一化與特征融合空間置棄的耦合堆疊可有效抑制多尺度目標(biāo)語義分割過擬合，提高模型精度和泛化性能。研究提出的模型及構(gòu)建的面向濱海生態(tài)環(huán)境監(jiān)管的多目標(biāo)語義分割數(shù)據(jù)集可為濱海區(qū)域生態(tài)修復(fù)、測繪和綜合治理提供決策支持。
• To improve the lack of deep learning dataset of coastal ecological scenes and low accuracy of multi-scale objectives semantic segmentation for remote sensing image classification, we take three types of coastal typical ecological supervision multi-scale objectives of mangrove, raft cultivation and pond aquaculture as research objects, constructs a benchmark dataset for coastal ecological supervision, improves the UNet feature fusion by integrating batch normalization and spatial dropout modules, and proposes a multi-scale deep convolutional semantic segmentation model. The model has an overall accuracy of 92% on the test set, a kappa coefficient of 0.87, and a mIoU of 82%. The experimental results show that the coupled stacking of batch normalization and feature fusion spatial dropout can effectively suppress multiscale objectives semantic segmentation overfitting and improve the model accuracy and generalization performance. The proposed model and the constructed semantic segmentation dataset for coastal ecological supervision can provide decision support for ecological restoration, mapping and comprehensive management in coastal areas.
• DOI： https://doi.org/10.35534/er.0402007
• 引用： 陳巖，楊曉彤，奚硯濤，等．面向濱海生態(tài)監(jiān)管的多尺度目標(biāo)語義分割研究［J］．環(huán)境與資源，2022，4（2）：48-61．