结合分层技术和SVM监督分类的植被专题信息提取  

Vegetation Identification of Remote Sensing Image Based on Hierarchical Clustering Analysis and SVM Supervised Classification

作　　者：刘炜[1] 

机构地区：[1]西藏民族大学信息工程学院,咸阳712082

出　　处：《软件导刊》2015年第10期160-162,共3页Software Guide

基　　金：国家自然科学基金项目(41361044);西藏民族大学青年学人项目(13myQP09)

摘　　要：比较最大似然法与"结合分层技术和SVM监督分类"方法,从神木县OLI图像上识别7种主要植被类型的精度。将神木县7种主要植被类型划入5个专题图层;对图像进行LBV变换后,通过阈值分割获取目标植被类型的概貌图像;以波段L、V、B作为有效特征进行面向对象分割和SVM监督分类,并在分类后执行开、闭运算操作,获得目标植被类型的精确提取结果;将该提取结果作为掩膜区域从原图上去除;重复上述过程依次处理5个专题图层,将各专题层提取结果叠加形成分类图;与最大似然法分类结果进行比较。结果表明,"结合分层技术和SVM监督分类"的方法能够有效降低OLI图像分类后的"椒盐效应",准确识别神木县7种植被类型,总体分类精度和Kappa系数分别为85.32%、0.796,较最大似然法分类结果分别提高了16.46%和17.93%。This research aims to seek out the most suitable vegetation classification method for Landsat OLI images by comparing supervised classification based on maximum likelihood and the method proposed in this paper. There are 5 the- matic image layers including 7 kinds of vegetation type in OLI images of Shenmu County. Firstly, OLI images were fused with panchromatic image after routine image preprocessing. Secondly, LBV transform were applied to OLI images. And then threshold segmentation in L-V feature space was carried out to get general cover information of target vegetation type. Thirdly, training samples？ set for target vegetation type and interference type were collected, and then object orien- ted segmentation, supervised classification based on SVM,opening-closing operation in mathematical morphology were car- ried out to get precise information of target vegetation type. Fourthly, precise information of target vegetation type was re- moved from original image for next vegetation type. Fifthly,classification results of 5 thematic image layers were stacking togethcr, therefore vegetative classification results of OLI image in Shenmu County was acquired. Assessing the classifica- tion results for method proposed in this paper and maximum likelihood, by overall classification accuracy and Kappa coeffi- cient as evaluation indexes. Results shows that： the overall accuracy and Kappa coefficient of classified image using the？ method proposed in this paper were 85.32% and 0. 796,with growth of 16.46% and 17.93% compared with classification image rising maximum likelihood. Meanwhile, removal of salt and pepper noise in classified images was more effective using method proposed in this paper.

关 键 词：监督分类 分层技术 OLI图像 面向对象分割 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]