| 摘要: |
| 【目的】 传统多尺度分割算法所获得的分割尺度多依靠经验确定，容易造成区域“过分割”和“欠分割”问题。文章利用高分辨率遥感影像地物间边缘信息突出的特点，提出了一种耦合边缘检测算子的多尺度自适应分割方法。【方法】 （1）首先利用Canny算子对遥感影像进行边缘提取，得到边缘梯度信息。（2）将初始多尺度分割结果与边缘梯度信息进行耦合处理，构建最优分割的评价指标选取地块的最适分割尺度。（3）将选取出的不同分割尺度结果进行合并，获得优化后的地块提取结果。【结果】 耦合边缘检测算子的多尺度自适应分割方法与传统的单一尺度分割方法相比，所得到的地块平均交并比提高了10.82%~27.11%，精度明显优于传统方法，能为耕地地块信息的精确提取提供参考。【结论】 耦合边缘检测算子的多尺度自适应分割方法在地块较为破碎的区域适用性较高，减少了区域的欠分割和过分割现象，能够保持地块区域的完整性，使得到的耕地地块更加符合真实地块轮廓。 |
| 关键词: 耕地地块提取 多尺度分割 Canny算子 最优分割尺度 |
| DOI：10.12105/j.issn.1672-0423.20240602 |
| 分类号: |
| 基金项目:国家重点研发计划“盐碱地土壤质量智慧监测技术”（2023YFD200140101）；国家自然科学基金“耕地利用规模的多尺度智能化遥感监测”（42071419）；中国农业科学院重大科技任务“盐碱地分区改良和综合利用技术集成示范”（CAAS-ZDRW202407） |
|
| A study of a multi-scale adaptive segmentation method with coupled edge detection operators |
|
Lai Linxiao1,2, Lu Miao1, Chen Dechao3, 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 Environmental Science and Engineering，Suzhou University of Science and Technology，Suzhou 215009，Jiangsu，China;3.School of Geographic Science and Surveying Engineering，Suzhou University of Science and Technology，Suzhou 215009，Jiangsu，China
|
| Abstract: |
| ［Purpose］ At present，the segmentation scale obtained by the traditional multi-scale segmentation algorithm is mostly determined by experience，which easily leads to over-segmentation and under-segmentation problems in the region.This study develops a multiscale adaptive segmentation method with coupled edge detection operators by taking advantage of the prominent edge information between features in high-resolution remote sensing images.［Method］ First，the Canny operator was utilized to extract the edges of the remote sensing image to obtain the edge gradient information.Then the initial multi-scale segmentation results were coupled with the edge gradient information，and the optimal segmentation index was constructed to select the most suitable segmentation scale of the parcel.Finally，the results of different segmentation scales were combined to obtain the optimized parcel extraction results.［Result］ Compared with the traditional single-scale segmentation method，the multi-scale adaptive segmentation method with coupled edge detection operator improved the average intersection and merger ratio of the obtained plots by 10.82 to 27.11 percent，and the accuracy was obviously better than that of the traditional method，which could provide a reference for the accurate extraction of cultivated land plot information.［Conclusion］ The multi-scale adaptive segmentation method with coupled edge detection operator has higher applicability in the region with more fragmented plots，reduces the under-segmentation and over-segmentation phenomenon of the region，and is able to maintain the integrity of the plot region，so that the obtained cultivated plots are more in line with the real plot contour. |
| Key words: cultivated land plot extraction multi-scale segmentation Canny operator optimal segmentation scale |
