基于无人机可见光影像的荒漠植被覆盖度提取研究

Study on Extraction of Desert Vegetation Coverage Based on UA V Visible Light Image

准确掌握荒漠植物生长状况是干旱地区荒漠化状态诊断的基础。本研究以典型荒漠区民勤为研究区,以无人机可见光影像为数据源,通过5种植被指数进行覆盖度信息提取,分析可见光植被指数在稀疏荒漠植被覆盖监测中的适用性,结果显示:1)荒漠植被(梭梭、白刺、红柳和伴生草本)各波段的像元值均表现为绿光波段像元值>红光波段像元值>蓝光波段像元值;2)RGRI和NGRDI中非植被的流沙和丘间地与植被数值范围存在交叉重叠,但EXG、NGBDI与VDVI影像中植被与非植被地物的指数值基本没有重叠交叉现象;3)从植被覆盖提取精度分析,植被信息提取准确率为EXG<RGBDI<NGBDI<VDVI,非植被信息提取正确率为EXG<RGBDI<NGBDI<VDVI,VDVI植被提取准确率达84.21%。因此,基于无人机可见光影像为数据源的情况下,荒漠植被与非植被在可见光的红光和蓝光波段容易区分,但在绿光波段不利于区分;EXG、NGBDI与VDVI更适用于荒漠植被信息的提取,在仅有可见光波段的无人机影像植被信息的提取中具有较好的适用性。

Accurate understanding the growth status of desert plants is the basis for diagnosing desertification in arid areas.This research takes Minqin,a typical desert region,as the research area,and uses the visible light shadow images of UAV as the data source.The coverage information is extracted through five vegetation indices,such as EXG,NGRDI,NGBDI,RGRI,and VDVI to analyze the applicability of the visible light vegetation index in sparse desert vegetation coverage monitoring.The results show that:1)The pixel values of desert vegetation(such as Haloxylon haloxylon,Nitraria tangutorum,Tamarix ramosissima and associated herbaceous plants)in each waveband are as follows:green band>red band>blue band;2)In RGRI and NGRDI,there is overlap between the values of vegetation and non-vegetationand(quicksand and interdune area),but in EXG,NGBDI and VDVI,the index values basically did not overlap and cross;3)Analyzing the accuracy of vegetation cover extraction,the accuracy of vegetation information extraction is EXG<RGBDI<NGBDI<VDVI,the accuracy of non-vegetation information extraction is EXG<RGBDI<NGBDI<VDVI,and the vegetation extraction accuracy of VDVI is 84.21%.Therefore,in the case of UAV visible light image,desert vegetation and non-vegetation are easy to distinguish in the red or blue bands of visible light,but not conducive to distinguish in the green band;EXG,NGBDI and VDVI are more suitable for the extraction of desert vegetation information,and have better applicability in the extraction of vegetation information from UA V images with only visible light bands.