摘要
【目的】耕地信息的提取和变化监测是遥感应用研究的热点之一,粤北地区是广东省主要的粮食基地,是业务管理部门进行耕地变化监测的重点地区。【方法】本文利用高分2号和哨兵2号遥感影像,结合研究区晚稻的物候期,通过提取植被指数、湿度指数、亮度指数、色彩指数并结合纹理特征进行多尺度分割,构建水田提取的多特征时序图像,运用随机森林法分别对晚稻生长期影像、晚稻收割后影像、时序指数差值图像和时序多特征差异图像进行分类,提取水田的面积,并对结果进行评估。【结果】(1)水田的提取精度在时序多特征差异图像的最高为0.98,与晚稻播种的面积差异最小为240.05 hm^(2),基于时序多特征差异图像的水田提取效果最好;(2)指数特征的差异需要利用作物的关键物候特征,并结合影像的多尺度分割,提高水田提取的准确程度;(3)随机森林方法在高维特征的数据分类时,具有较快的运算速度和较高的分类精度。【结论】本文的研究是对高分辨遥感在耕地信息快速准确更新的方法探索,可为耕地管理业务提供技术支撑。
【Objective】The extraction and change monitoring of cultivated land information is one of the hotspots of remote sensing application research.Northern Guangdong is the main grain base of Guangdong Province,and is the key area of cultivated land change monitoring by business management departments.【Method】In the present paper,using GF-2 and Sentinel-2 remote sensing images,combined with the phenological period of late rice in the study area,the vegetation index,humidity index,brightness index,color index and texture features were extracted for multi-scale segmentation,and the multi-scale image of paddy field extraction was constructed.The index difference image and time series multi feature difference image were classified to extract the paddy field area and evaluate the results.【Result】(i)The highest extraction accuracy of paddy field in time series multi feature difference image was 0.98,and the minimum difference between paddy field area and late rice sowing area was 240.05 hm^(2),and the best paddy field extraction effect was based on time series multi feature difference image;(ii)The difference of index feature needed to use the key phenological characteristics of crops and combine with multi-scale segmentation of image to improve the accuracy of paddy field extraction;(iii)The machine forest method has faster operation speed and higher classification accuracy in data classification with high dimensional features.【Conclusion】The research of the paper is to explore the method of high-resolution remote sensing in rapid and accurate updating of cultivated land information,which can provide support for the application of cultivated land management business.
作者
王卫
朱明帮
陈晓远
胡月明
林昌华
WANG Wei;ZHU Ming-bang;CHEN Xiao-yuan;HU Yue-ming;LIN Chang-hua(College of Natural Resources and Environment,South China Agricultural University,Guangdong Guangzhou 510642,China;Guangdong Key Laboratory for Land Use and Consolidation,Guangdong Guangzhou 510642,China;Guangdong Land Information Engineering Technology Research Center,Guangdong Guangzhou 510642,China;Yingdong College of Biology and Agriculture,Shaoguan University,Guangdong Shaoguan 512005,China;North Guangdong Soil and Land Research Center,Shaoguan University,Guangdong Shaoguan 512005,China)
出处
《西南农业学报》
CSCD
北大核心
2021年第10期2223-2230,共8页
Southwest China Journal of Agricultural Sciences
基金
国家重点研发计划(2018YFD1100103)
广东省自然科学基金(2018A030307075)
2018年广东省科技创新战略专项资金(粤科函规财字[2018]1523号)
韶关市科技计划项目(201644)。
关键词
水田提取
随机森林
时序多特征
物候差异
粤北地区
Paddy field extraction
Random forest
Time series multi feature difference
Phenological difference
Northern Guangdong area
