摘要
针对传统的大规模草地牧草的识别,不仅浪费大量的人力、物力,还浪费大量的时间和经费。该文提出了一种基于高光谱成像采集系统的草地动态监测新手段,利用采集到的可见-近红外光谱(范围400~1 000nm)图像和光谱信息进行自动分类。该方法主要包括预处理、特征波段提取以及分类识别3个步骤。(1)利用ENVI(4.7)提取图片的感兴趣区域光谱数据,由于存在大量的数据冗余以及外界的噪声干扰等因素,因此采用多元散射校正去除散射,增强与成分含量相关的光谱吸收信息。(2)采用连续投影选取出13个特征波段消除数据冗余。(3)采用支持向量机对选取的特征波段进行分类,分类识别率可达100%。结果表明,采用高光谱成像技术对野外牧草种类的无损识别是可行的,SPA提取光谱特征波段及SVM进行判别野外田间牧草种类取得较好的效果。
According to the traditional identification of large-scale pasture,it not only wastes a lot of manpower and material resources,but also waste a lot of time and money.In this paper,a new approach to dynamic monitoring of grassland based on hyperspectral imaging acquisition system was proposed,in which the captured visible and near infrared spectroscopy（range of 400~1 000 nm）images and spectral information could be automatically classified.Three main steps were involved in the method including preprocessing,feature band extraction and classification.Initially,the region of interest（region of interest,ROI）of the image is extracted by ENVI（4.7）.Because of the large amount of data redundancy and external noise interference,successive scattering correction（support vector machine,SPA）is used to remove the scattering and strengthen the correlated spectral absorption information.Subsequently,continuous projection was applied to select 13 feature bands to eliminate data redundancy.Finally,the support vector machine（support vector machine,SVM）was used for classification of the selected characteristic bands,yielding a classification accuracy of 100%.The results showed that the hyperspectral imaging technique is feasible for non-destructive identification of forage species,and the proposed method using SPA and SVM to extract spectral characteristics and classification had a preferable classification result for forage species in wild field.
作者
武林
马玉宝
潘新
闫伟红
段俊杰
WU Lin;MA Yubao;PAN Xin;YAN Weihong;DUAN Junjie(College of Computer and Information Engineering,Inner Mongolia Agricultural University,Hohhot 010020,China;Institute of Grassland Research of CAAS,Huhhot 010010,China)
出处
《测绘科学》
CSCD
北大核心
2018年第9期52-57,共6页
Science of Surveying and Mapping
基金
国家自然科学基金项目(61562067
31302017)
