| 摘要: |
| 【目的】 针对地表覆盖变化受干扰因素影响而产生伪变化的问题，提出一种基于遗传规划（Genetic Programming，GP）的变化检测方法，以提高土地管理、灾害评估和生态监测中地表覆盖变化检测的准确性和实时性。【方法】 文章以山东省阳谷县为研究区，基于研究区两幅Landsat-8 OLI影像，采用GP算法进行种群初始化、适应度评估和种群进化，进而进化出最佳个体用于地表覆盖变化检测，并将GP检测结果与变化向量分析、主成分分析和相关系数法这3种传统遥感变化检测方法进行比较。【结果】 基于GP的变化检测方法获得的总体精度为95.55%，Kappa系数为0.91，变化和未变化像元的错分率分别为3.37%和5.50%，漏分率分别为5.58%和3.32%，优于3种传统变化检测方法。【结论】 基于GP的变化检测方法可以自动选择地表覆盖敏感光谱波段构建变化强度树，并自动集成阈值和变化强度树，避免了手动设置阈值带来的主观干扰。 |
| 关键词: 变化检测 遗传规划 土地覆盖 遥感 |
| DOI：10.12105/j.issn.1672-0423.20230404 |
| 分类号: |
| 基金项目:国家重点研发计划“全球变化对粮食产量和品质的影响研究”（2019YFA0607401）；国家自然科学基金项目“耕地规模化利用的多尺度智能遥感监测方法研究”（42071419）；国家自然科学基金项目“面向复杂少样本图像分类的遗传规划算法研究”(62376253）；中国农业科学院科技创新工程“盐碱地产能提升关键技术与集成示范”（CAAS-ZDRW202201） |
|
| Automatically evolving land cover change detector using genetic programming |
|
Zhang Shengnan1, Lu Miao1, Bi Ying2, Wen Caiyun1
|
|
1.State Key Laboratory of Efficient Utilization of Arid and Semi-arid Arable Land in Northern China /Institute of Agricultural Resources and Regional Planning，Chinese Academy of Agricultural Sciences，Beijing 100081，China;2.School of Electrical and Information Engineering，Zhengzhou University，Zhengzhou 450001，Henan，China
|
| Abstract: |
| 【Purpose】 To address the issue of pseudo changes caused by interference factors in land cover changes，a change detection approach based on genetic programming（GP）is proposed to improve the accuracy and real-time performance of land cover change detection in land management，disaster assessment，and ecological monitoring.【Method】 This study taked Yanggu County，Shandong Province as the research area，and based on 2 Landsat-8 OLI images for GP-based population initialization，fitness evaluation，and population evolution. The best individuals evolved by GP were used for land cover change detection. The GP detection results were compared with traditional remote sensing change detection methods including change vector analysis（CVA），principal component analysis（PCA），and correlation coefficient（CC）.【Result】 The overall accuracy obtained by GP was 95.55% and the Kappa coefficient was 0.91，the commissions of changed and unchanged pixels were 3.37% and 5.50%，the omissions were 5.58% and 3.32%，respectively. These results outperform the 3 traditional change detection methods.【Conclusion】 The proposed approach obtained by GP can automatically select sensitive spectral bands of land cover to construct a change intensity tree，and automatically integrate threshold and change intensity trees，avoiding subjective interference caused by manually setting thresholds. |
| Key words: change detection Genetic Programming land cover remote sensing |
