摘要
桉树生长速度快、用途广,具有良好的经济效益、社会效益和生态效益。但桉树的种植、砍伐与促长之间周期交替过于频繁,传统的林业调查不仅会受各种主客观因素的影响,而且获取信息的可信度较低。文章以Rapi Eye影像为数据源,结合GPS调查手段,运用e Cognition平台,以光谱和纹理相结合的面向对象的分析方法,并创建知识规则集,对桉树林信息进行提取。实验表明,传统基于像素的监督分类方法总体精度为68.83%,Kappa系数为0.67。而基于以光谱纹理、创建知识规则集相结合的面向对象分析方法,桉树信息提取的总体精度达到82.12%,Kappa系数为0.80,该提取方法可获得更好的分类质量、效率与精度。
Eucalyptus growth speed, wide application, has the good economic benefit, social benefit and ecological benefit. But the eucalyptus trees planting, cutting and growth cycle, alternating between too frequently, the traditional forestry investigation will not only influenced by various subjective and objective factors, the credibility of and access to information is low. RapiEye images as data source, this paper combined with GPS survey method, combining with spectrum and texture objectoriented analysis method, rule set and create knowledge, to extract eucalyptus forest information. Experiments show that the traditional supervised classification method based on pixers overall accuracy is 68.83%, the Kappa coefficient is 0.67. And based on texture spectrum, create knowledge rule sets the object-oriented analysis method of combining the overall precision of eucalyptus information extraction of 82.12%, Kappa coefficient is 0.80, the extraction method can get better classification quality, efficiency and accuracy.
出处
《大众科技》
2016年第4期16-23,共8页
Popular Science & Technology
基金
广西自然科学基金项目(2014GXNSFAA118293)
关键词
光谱特征
纹理特征
桉树林
遥感信息
提取方法
Spectral characteristics
texture feature
eucalyptus forest
remote sensing information
extraction method
