摘要
危险性林业有害生物每年都会造成大面积的严重的森林灾害发生,给国民经济和林业生态环境带来了巨大损失,及时对其发生发展情况进行监测具有重要意义。本文旨在探讨利用高空间分辨率遥感影像识别红叶松树的可行性,从而为单株木灾害监测与防治提供一种快速有效的技术途径。采用World View-2多光谱数据,基于二次型分类器,研究了红叶松树的识别方法。通过Fisher线性判别分析发现,World View-2的红、绿、蓝3个波段蕴含着丰富的红叶松树信息。据此,基于全部8个波段和红、绿、蓝3个波段分别进行了分类,分类精度基本相同。结果表明,仅利用红、绿、蓝3个波段也可以完成红叶松树的识别,从而有效地实现了数据降维,更为我国"高分二号"数据的应用开辟了新领域。由于林相条件的差异性和复杂性,遥感数据分辨率的选择对于单木灾害的识别非常重要,研究发现,2 m分辨率多光谱遥感数据非常适合红叶松树的提取,且从数据时相上来说,最好选取6—9月份的遥感影像,以减少其他变色植被的干扰,提高监测精度和工作效率。
Dangerous pests are severe forest disasters,and they occur every year in large areas and cause great damage to forest ecology and tremendous loss to economy. It has great significance to detect the occurrence of the hazard in time and to monitor its spreading process. The purpose of this study is to probe into the possibility to recognize red-attack pine trees with high resolution imagery,so as to provide a quick and efficient methodology for monitoring,prevention and control of fatal forest pests. The recognition of red-attack pine trees was studied based on the World View-2 multi-spectral data using the analysis method of the quadratic classifier. Through the Fisher determination analysis,we found that the red,green and blue bands have rich information of red-attack pine trees. Based on this discovery,we first conducted classification with all the eight bands,then with only the red,green and blue bands,and did comprehensive analysis. Results show that the accuracies of the two classifications are nearly the same,which means that the red,green and blue bands are enough to recognize the red-attack pine trees,so that the dimension of raw imagery can be reduced. This also contributes to the new application areas of GF-2 imagery. Because of the complexity of forest phases,the selection of image resolution is very important in the recognition of individual trees. Researches show that the 2-m resolution remote sensingdata can well be used in extracting the red-attack pine trees. Moreover,it is better to choose data in the vegetation growing season,so that the discoloration mixture phenomenon can be reduced and the accuracy and efficiency of monitoring can be increased.
出处
《北京林业大学学报》
CAS
CSCD
北大核心
2015年第11期33-40,共8页
Journal of Beijing Forestry University
基金
国家重大专项(21-Y30B05-9001-13/15)
关键词
WorldView-2多光谱影像
红叶松树
变色立木
遥感监测
World View-2 multispectral imagery
red-attack pine trees
discolored individual trees
remote sensing monitoring
