基于中高分影像的丘陵区柑橘种植地块提取

Extraction of citrus planting plots based on medium-high different images

【目的】利用多源中高分影像探索广西富川县柑橘种植地块尺度遥感监测方法,为实现全区柑橘信息精准监测提供技术参考。【方法】以高分辨率影像和二调数据为基底,更新图斑形态边界,生成完整且稳定的地块数据;分别面向农作物、果园、林地地块对象,综合中高分影像光谱信息、纹理信息及时间序列影像特征,利用支持向量机分类器迭代识别柑橘信息;基于DEM计算柑橘地块实际面积。【结果】通过分析不同地物影像特征曲线发现,传统指数NDVI(1、4、12月)、RVI(9—12月)、LSWI(1、10和12月)及红边指数NDRE1(1—2月、10—12月)、MTCI(1、2、10和12月)、PSSRa(9—12月)、MCARI2(1月、9—12月)对于识别柑橘作物具有敏感性。本研究提取柑橘信息精度为93.4%,比前人柑橘作物提取成果精度更高、尺度更精细。研究区柑橘地块种植于坡地面积占20.26%,在DEM支持下计算柑橘种植总面积为18643.15 ha,比采用投影面积测量方法增加148.49 ha,一定程度上消除地形对面积统计的影响。提取结果表明富川县柑橘主要集中在麦岭镇、富阳镇、葛坡镇、福利镇、朝东镇等,而莲山镇、白沙镇、古城镇种植柑橘较少。【建议】为解决广西立地条件的难题,建议以低频高分影像更新地块边界,在地块数据基础上综合利用多星多时相中高分影像(尤其具有红边波段影像)提取作物信息,针对地形起伏较大区域,建议采用高精度DEM计算作物面积。

【Objective】Using multi-source medium and high score images to explore the remote sensing monitoring method of citrus planting plots in Fuchuan,Guangxi,and provide technical reference for the realization of accurate monitoring of citrus plot information in the whole region.【Method】Based on high-resolution images and two-tone data,the morphological boundary of the patch was updated to generate complete and stable plot data. For crops,orchards,and woodland plot objects,the spectral information,texture information,time series image features of the mid-high score image were integrated,and the support vector machine classifier was used to iteratively identify the citrus information. DEM was calculated based on a true citrus land area.【Result】Analysis of the characteristic curve indicated that the traditional indexes NDVI(January,April,and December),RVI(September to December),LSWI(January,October,and December)and the red edge indexes NDRE1(January to February,October to December),MTCI(January,February,October,and December),PSSRa(September to December),MCARI2(January,September to December)were sensitive to the identification of citrus crops. The accuracy of extracting citrus information in this study was 93.4%,which was higher than the existing citrus crop extraction results and had a finer scale. The citrus plot in the study area accounted for 20.26% of the sloping land area. Under the action of DEM,the citrus planting area was calculated to be 18643.15 ha,which was an increase of 148.49 ha compared with the projection area measurement method. To a certain extent,the influence of terrain on area statistics was eliminated. The extraction results showed that the citrus in Fuchuan County were mainly concentrated in the towns of Mailing,Fuyang,Gepo,Fuli,Chaodong,while the towns of Lianshan,Baisha and Gucheng had less citrus planting.【Suggestion】In order to solve the problem of site conditions in Guangxi,it is recommended to update the plot boundaries with low-frequency and high-resolution images,and comprehensively use multi-star and multi-temporal mid-and high-resolution images(especially images with red edge bands)to extract crop information based on the plot data. For areas with large terrain fluctuations,it is recommended to use high-precision DEM to calculate the crop area.