基于多时相遥感影像的北京平原人工林树种分类

发布时间：2018-02-20 20:37

本文关键词： 多时相影像面向对象最优分割尺度特征筛选平原林地树种分类　出处：《北京工业大学学报》2017年05期 　论文类型：期刊论文

【摘要】：为解决传统遥感分类方法区分平原人工造林地树种难度较大的问题,利用4个不同时相的高空间分辨率卫星影像,基于ESP计算方差变化率并结合目视解译获取影像的最佳分割尺度;通过相关系数法筛选构建的特征,进行面向对象的多时相影像和单时相影像分类,并与基于像元分类方法进行对比分析.结果表明:基于多时相影像各类别分类精度为64%,高于单时相分类精度(51%);面向对象KNN方法的分类精度优于SVM和MLC分类方法,两者精度分别为49%和43%.在树种丰富且分布复杂的平原造林林地景观中,利用多时相遥感数据,采用面向对象分类方法用于树种精细分类更具优势.
[Abstract]:In order to solve the problem that it is difficult for the traditional remote sensing classification method to distinguish the tree species from the plain artificial woodland, four high spatial resolution satellite images with different phases are used. Based on ESP, the best segmentation scale of image is obtained by combining the variance change rate and visual interpretation, and the features constructed are screened by correlation coefficient method, and the object oriented multitemporal image and single phase image classification are carried out. The results show that the classification accuracy of each classification based on multitemporal image is 64, which is higher than that of single phase classification, and the accuracy of object oriented KNN is better than that of SVM and MLC. The precision of the two methods is 49% and 43 respectively. In the forestland landscape with abundant species and complicated distribution, it is more advantageous to use the multi-temporal remote sensing data and the object-oriented classification method in the fine classification of tree species.
【作者单位】：北京农业信息技术研究中心;四川理工学院四川省院士(专家)工作站;中国科学院遥感与数字地球研究所数字地球重点实验室;
【基金】：国家自然科学基金资助项目(41571423) 中国科学院青年创新促进会资助项目
【分类号】：S771.8;TP751

【相似文献】