摘要
为了提高遥感影像分类精度,对传统的非监督分类、监督分类和专家分类进行机理分析,提出将影像中的纹理信息作为专家知识改进分类精度的技术方案。以胶州市QuickBird遥感影像作为试验数据,基于ERDAS IMAG-INE 8.6软件平台,对非监督分类、监督分类和专家分类进行实验数据比较分析,实验数据表明改进的专家分类方法分类精度最高,由于纹理信息参与专家分类,可较好地解决“同谱异物”和“同物异谱”对分类的干扰,优化分类后的影像,提高信息提取的准确度。
To improve remote sensing image classification accuracy, analysis is made on the mechanism of the traditional unsupervised classification, supervised classification and expert classification. It puts forward the technological program of using the image texture information as the expert knowledge to improve the classification accuracy. Using the QuickBird remote sensing image of Jiaozhou city as the trial data, based on the software platform of ERDAS IMAGINE 8.6, the test data indicates that the classification accuracy of advanced expert classification method is the best, because that texture information participates in the expert classification, which better solves the classification disturbance of different objects with the same spectrum and the same objects with the different spectrums.
出处
《测绘工程》
CSCD
2007年第3期31-34,39,共5页
Engineering of Surveying and Mapping
基金
辽宁省自然科学基金资助项目(20042175)
关键词
影像分类
专家分类
知识库
纹理信息
image classification
expert classification
knowledge base
texture information
