一种基于光谱归一化下的植被覆盖度反演算法  

A Kind of Vegetation Cover Fraction Retrieval Algorithm based on Spectral Normalization Frame

作　　者：段金亮[1] 王杰[1] 张婷[1] Duan Jinliang,Wang Jie,Zhang Ting(College of Land and Resources ,China West Normal University ,Nanchong 637009,China)

机构地区：[1]西华师范大学国土资源学院,四川南充637009

出　　处：《遥感技术与应用》2018年第2期252-258,共7页Remote Sensing Technology and Application

基　　金：国家自然科学基金面上项目(41671220);西华师范大学博士科研启动基金(412546;412547);四川省教育厅自然科学重点项目(17AZ0387);2016年西华师范大学大学生创新创业自然科学论文项目(cxcy2016033);四川省大学生创新创业训练项目(201710638057)

摘　　要：针对常规混合像元分解算法在植被覆盖度遥感反演中存在的端元变化误差及运算效率的问题,以两个不同类型植被覆盖下地区的TM影像数据为基础,提出了一种基于光谱归一化框架下的协同稀疏回归的植被覆盖度反演算法,并针对多种地表类型下的植被覆盖度反演试验,与常用的像元二分法模型进行对比分析。试验结果表明:对影像与端元组进行归一化后,有效地降低了它们的异质性,从而提高了反演精度,同时,该算法获取的植被覆盖度相对像元二分法具有更高的精度。Vegetation cover fraction is an important control factor in the process of simulating surface vegetation transpiration,soil water evaporation and vegetation photosynthesis.Based on the TM image data of two different types of vegetation cover,a collaborative sparse regression algorithm based on the spectra normalization framework is proposed to retrieve the vegetation cover fraction,which solves the problems such as the error of the endmember variability and the efficient of the algorithm arisen from many spectral mixture analysis algorithms used to retrieving vegetation cover fraction.And also by contrast to the dimidiate pixel algorithm,the accuracy of the algorithm is indicated.The experimental results show that the normalization of the image and endmenbers can effectively reduce their heterogeneity and improve the retrieval precision and the algorithm has higher accuracy than the dimidiate pixel algorithm.

关 键 词：光谱归一化 植被覆盖度 TM与ALI影像 像元二分法 协同稀疏回归算法 端元变化 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]