基于高分二号卫星影像的粤北地区香芋遥感识别研究  

Study on Remote Sensing Identification of Taro in Northern Guangdong Based on GF-2 Satellite Image

作　　者：王卫[1,2,3,4] 李敬 陈晓远 高琳[1,2] WANG Wei;LI Jing;CHEN Xiaoyuan;GAO Lin(Yingdong College of Biology and Agriculture,Shaoguan University,Shaoguan 512005,China;Yuebei Soil and Land Research Center of Northern Guangdong,Shaoguan University,Shaoguan 512005,China;College of Natural Resources and Environment,South China Agricultural University,Guangzhou 510642,China;Land Science and Technology Center,South China Agricultural University,Guangzhou 510642,China)

机构地区：[1]韶关学院英东生物与农业学院,广东韶关512005 [2]韶关学院粤北土壤土地研究中心,广东韶关512005 [3]华南农业大学资源环境学院,广东广州510642 [4]华南农业大学土地科技中心,广东广州510642

出　　处：《广东农业科学》2020年第6期126-133,共8页Guangdong Agricultural Sciences

基　　金：广东省自然科学基金(2018A030307075);广东省科技创新战略专项资金(粤科函规财〔2018〕1523号);韶关市科技计划项目(201644);韶关学院校级课题项目(2015262)。

摘　　要：【目的】高分辨率遥感影像为农作物监测提供高精度的支撑,香芋作为粤北地区的特色作物,是国家地理标志产品,对其监测有助于加强管控和调控。【方法】选择香芋关键物候期的高分二号卫星遥感影像,提取归一化植被指数、归一化差异水体指数、纹理信息,构建融合多特征光谱纹理影像,比较多种组合影像,采用支持向量机作为分类器,对香芋的识别精度进行分析。【结果】融合多特征光谱纹理影像的香芋识别精度最高,总体精度达到96.04%,对香芋的识别精度达到95.30%,比多光谱影像分类精度分别提高5%和6.8%,是多光谱全色融合影像分类精度提升幅度的2倍,且各类地物边界轮廓清晰,图像平滑,细碎图斑很少。【结论】高分二号影像是识别粤北地区香芋的理想数据源,分类精度较高,能够满足农作物监测的需求,能为制定病虫害防治措施,调节种植结构提供支持。【Objective】The high-resolution remote sensing image provides high-precision support for crop monitoring.As the characteristic crop in northern Guangdong province,taro is a national product of geographical indication,and the monitoring of which is helpful to strengthen control and regulation.【Method】In this paper,the GF-2 remote sensing image of the key phenology of taro was selected,the normalized vegetation index,normalized difference water index and texture information were extracted,the multi-feature spectral texture images were constructed and multiple image combinations were compared,and the support vector machine was used as classifier to analyze the identification accuracy of taro.【Result】The identification accuracy of taro integrated with multi-feature spectral texture image was the highest,with the overall accuracy of 96.04%,and the classification accuracy of taro was 95.30%,which was 5%and 6.8%higher than that of multispectral image,respectively,with twice compared with the improvement of classification accuracy of multispectral panchromatic fusion image.And the boundary outlines of all kinds of ground objects were clear,with smooth image and few fine spots.【Conclusion】GF-2 is an ideal data source for identifying taro in northern Guangdong province,with high classification accuracy,which can meet the needs of crop monitoring,provide support for the formulation of pest prevention and control measures and the adjustment of planting structures.

关 键 词：遥感识别 纹理信息 多光谱特征 粤北地区 香芋 

分 类 号：S632.3[农业科学—蔬菜学] S127[农业科学—园艺学]