利用多时相特征的落叶松人工林分类

基于多时相遥感影像研究落叶松Larixgmelini人工林季相特性，采用多种分类方法提取落叶松人工林空间分布，以期得到适用落叶松人工林提取的手段和方法。利用多时相Landsat8影像，在分析落叶松人工林季相和物候特性的基础上，综合森林资源二类调查成果和样地数据，分别采用最大似然法、支持向量机法、光谱角法和k最近邻法提取研究区的落叶松人工林信息，并进行精度验证。落叶松人工林具有明显的植被光谱特征和季相特性，在近红外波段各树种类别光谱差异最明显，秋季影像对落叶松人工林信息提取效果较好。利用多时相数据对落叶松人工林信息提取精度明显优于单时相数据的提取精度，多时相Landsat8影像能使落叶松人工林分类精度达86%以上，其中光谱角法最好，精度为88.3463%。

Studying the phenological and seasonal characteristics of Larix gmelini plantations using multi-temporal remote sensing images, we adopted multiple classification methods to extract the spatial distribution of such plantations, aiming to acquire applicable approaches for Larix gmelini plantation extraction. Using multi-temporal Landsat 8 images, and based on the analysis of the seasonal and phenological traits of Larix gmelini plantations combined with the outcomes of the second-class forest resource inventory and field plot data, we applied Maximum Likelihood Classification (MLC), Support Vector Machine (SVM), Spectral Angle Mapper (SAM), and k-Nearest Neighbor (k-NN) to extract information on Larix gmelini plantations in the study area, and conducted accuracy assessment. Larix gmelini plantations possess distinct vegetation spectral features and seasonal characteristics; the spectral differences among tree species are most prominent in the near-infrared band, and autumn images deliver better performance for extracting plantation information. The extraction accuracy of Larix gmelini plantation information using multi-temporal data is significantly superior to that using single-temporal data. Multi-temporal Landsat 8 images can achieve a classification accuracy of over 86% for Larix gmelini plantations, among which the Spectral Angle Mapper performs the best, with an accuracy of 88.3463%.

搜集汇总

背景与挑战

背景概述

该数据集基于多时相Landsat8影像，研究落叶松人工林的季相特性，并应用多种分类方法提取其空间分布。结果表明，多时相数据能显著提升分类精度，最高可达88%以上，其中光谱角法表现最佳。