水土保持措施识别与提取方法的研究进展
DOI:
CSTR:
作者:
作者单位:

华中师范大学

作者简介:

通讯作者:

中图分类号:

基金项目:

国家自然科学基金面上项目(42377354);教育部“春晖计划”合作科研项目(202200199);水利部水网工程与调度重点实验室开放研究基金(QTKS0034W2328)。


Research progress on identification and extraction methods of soil and water conservation measures
Author:
Affiliation:

Central China Normal University

Fund Project:

The National Natural Science Foundation of China (General Program, Key Program, Major Research Plan)

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    水土保持措施类型及其配置模式复杂繁多,准确识别与精细化提取水土保持措施详细配置信息是获取水土保持措施因子值的基础。水土保持措施信息获取方式主要有传统的野外调查、卫星遥感影像、无人机近景摄影等,其识别与提取方法主要包括目视解译、传统的机器学习、面向对象分类方法以及深度学习模型。本文通过梳理国内外水土保持措施识别与提取方法的研究成果,总结存在的不足并提出研究展望;在语义分割中未来的特征融合与多模态学习、弱监督与半监督学习、集成学习和元学习等均可以运用到水土保持措施提取中;当前对于水土保持耕作措施识别与提取的成果鲜见报道,而农业实践中耕作措施较常见,后续应加强耕作措施识别提取的研究;人工智能结合大数据技术是未来水土保持措施信息高效精准识别与提取的发展方向,且需要进一步研究采用半监督和弱监督学习、语义分割和实例分割方法,结合多模态学习、小样本标签等方法进行水土保持点、线状工程措施及植物措施和耕作措施信息的提取,从而完善各类水土保持措施的信息提取方法,为准确获取水土保持措施因子值及核算水土保持碳汇能力提供支撑。

    Abstract:

    The types of soil and water conservation measures and their configuration modes are complicated. Accurate identification and fine extraction of detailed configuration information of soil and water conservation measures are the basis for obtaining the factor values of soil and water conservation measures. The information acquisition methods of soil and water conservation measures mainly include traditional field surveys, satellite remote sensing images, and UAV close-range photography. The identification and extraction methods mainly include visual interpretation, traditional machine learning, object-oriented classification methods, and deep learning models. This paper summarizes the existing shortcomings and puts forward the research prospect by combing the research results of soil and water conservation measures identification and extraction methods at home and abroad. In semantic segmentation, future feature fusion and multimodal learning, weak supervision and semi-supervised learning, ensemble learning and meta-learning can be applied to the extraction of soil and water conservation measures. At present, there are few reports on the results of identification and extraction of soil and water conservation tillage measures, while tillage measures are common in agricultural practice, and the research on identification and extraction of tillage measures should be strengthened in the future. Artificial intelligence combined with big data technology is the development direction of efficient and accurate identification and extraction of soil and water conservation measures information in the future, and it is necessary to further study the semi-supervised and weakly supervised learning, semantic segmentation and instance segmentation methods, combined with multi-modal learning, small sample label and other methods to extract the information of soil and water conservation points, linear engineering measures, plant measures and tillage measures, so as to improve the information extraction methods of various soil and water conservation measures, and provide support for accurately obtaining the factor values of soil and water conservation measures and accounting the carbon sink capacity of soil and water conservation.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2024-03-22
  • 最后修改日期:2024-04-29
  • 录用日期:2024-06-12
  • 在线发布日期: 2024-07-18
  • 出版日期:
文章二维码