基于特征优选的面向对象毛竹林分布信息提取  

Mapping of Moso Bamboo Forest Using Object-Based Approach Based on the Optimal Features

作　　者：高国龙[1] 杜华强[1] 韩凝[1] 徐小军[1] 孙少波[1] 李雪建[1] 

机构地区：[1]浙江农林大学环境与资源学院浙江省森林生态系统碳循环与固碳减排重点实验室,临安311300

出　　处：《林业科学》2016年第9期77-85,共9页Scientia Silvae Sinicae

基　　金：浙江省杰出青年科学基金项目(LR14C160001);国家自然科学基金项目(31300535;31370637;61190114);国家林业局"948"项目(2013-4-71);浙江省自然科学基金项目(LQ13C160002);浙江省本科院校中青年学科带头人学术攀登项目(pd2013239);浙江农林大学农林碳汇与生态环境修复研究中心预研基金项目

摘　　要：【目的】提出一种基于Relief F特征优选的面向对象分类方法,为解决面向对象森林资源遥感分类提供参考。【方法】以SPOT5高分辨率遥感影像为数据源,以浙江省安吉县山川乡为研究区,通过影像分割,选取8个地物类别的370个对象样本,并设置SPOT5影像每个波段的8个灰度共生矩阵纹理、每个波段及NDVI的平均值和标准差等42个对象特征。利用Relief F算法对设置的42个对象特征进行优选,采用面向对象的最近邻方法提取研究区毛竹林分布信息。为了比较基于最优特征的面向对象的分类结果,另采用CART决策树方法在相同的分割参数和训练样本前提下,通过样本构建决策树分类规则,对研究区进行分类并提取竹林信息。【结果】1)通过Relief F特征优选方法对分类特征进行优选,大幅提高了毛竹林样本的分类精度,与特征优选前相比,毛竹林样本分类精度由68%提高到88%,优选的红波段均值、绿波段均值、红波段均质纹理、红波段熵纹理和NDVI植被指数均值5个特征能够精确地提取研究区毛竹林分布信息,其用户精度和生产者精度分别达到97%和95%;2)基于CART决策树面向对象的研究区毛竹林用户精度和生产者精度均低于基于最优特征的最近邻分类结果,主要原因是CART决策树中毛竹林、针叶林和阔叶林之间的误分相对较高。【结论】Relief F算法特征优选时注重特征的分类能力,筛选的特征参与面向对象分割提取的毛竹林分布信息高于同类研究,可为面向对象多尺度分割森林资源遥感分类时特征的选取提供一个更为科学合理的方法。【Objective】Object-based classification method provides a new way for classifying remote sensing image of high spatial resolution because it can synthetically use spectral feature,geometric feature,texture feature,and so on.However,increase in the number of features will lead to the emergence of the＂dimension disaster＂,complicate operation,and decline in speed. It also leads to a decrease in classification accuracy under the limited training samples. In order to solve the problem in the feature space selection during object-based classification,an optimal features selection method based on Relief F was proposed in this study. 【Method 】The method gave a weight to each feature according to the separability of features of training samples,and selected the features which are important to samples classification through analyzing the correlation between features. Taking Shanchuan town in Anji County in Zhejiang Province as the study area,370 object samples for eight classes were selected. A total of 42 object features were selected,including mean and standard deviation of normalized difference vegetation index（ NDVI）,each band of SPOT5 image and their gray level cooccurrence matrix textures（ GLCM）. Based on the optimal features selected from the 42 object features using the Relief F algorithm,nearest neighbor algorithm in object-based classification method was used to extract the distribution of the bamboo forest in the study area. Moso bamboo forest information extracted by object-based classification based on optimalfeatures was compared with the results from classification and regression tree（ CART） decision tree algorithm in objectbased classification method,under the same segmentation parameters and training samples. 【Result 】1） By using the Relief F algorithm,the classification accuracy of bamboo forest samples has been greatly improved. The accuracy of moso bamboo samples was increased from 68% to 88%. Mean object value of red band as well as green band,mean component of GLCM in red

关 键 词：毛竹林 RELIEFF算法 特征优选 面向对象 SPOT5遥感影像 

分 类 号：S757[农业科学—森林经理学]