基于多特征空间的遥感专题信息自动提取方法研究

【作者】 杨桄；

【导师】 刘湘南；

【作者基本信息】 东北师范大学 ， 自然地理学， 2003， 硕士

【摘要】 遥感图像是探测地物目标综合信息的最直观、最丰富的载体，直接从遥感信息中自动提取地形和专题信息是当前遥感和地理信息系统技术面临的一个迫切而又复杂的问题。随着航天遥感信息获取技术的飞速发展，从遥感信息中自动提取专题信息已成为遥感信息生产流程中的瓶颈环节。采用基于光谱的特征提取和分类、单个目标图形特征提取等方法对遥感影像的自动解译已经研究了很多年，虽然取得了一些进展，但解译效果还达不到预期要求。本文研究了地物多种识别特征的遥感信息表达，提出了基于多特征空间的遥感专题信息自动提取的研究方法，将地物在地理空间和影像空间的多种特征空间分为光谱特征空间、形状特征空间、干扰特征空间和区域地学特征空间，通过对多特征空间的数学描述和计算机处理实现遥感专题信息的自动提取。通过对研究区玉米专题信息监督分类的结果评估，结合对玉米地分布影响因子的分析，提出了基于多特征空间的玉米专题信息自动提取模型，建立了多特征专家知识库并进行了TM影像专题信息自动分类。从分类结果与目视解译结果的比较中发现，基于多特征空间的专题信息自动提取方法可以提高专题信息的提取效率，对未来实现遥感影像的智能解译是一种研究方法的探索，具有一定的创新性和理论意义。全文共分为四部分，第一章分析了遥感影像自动解译的研究现状和发展趋势，阐明了本文的研究内容、研究意义、研究目的和研究方法。第二章阐述了地表物体多特征的遥感信息表达方式，提出了基于多特征空间的专题信息自动提取的方法，构建了基于多特征空间的专题信息自动提取模型。第三章以玉米地为典型专题信息，通过建立研究区玉米的多特征空间来实现自动提取，并与玉米监督分类的结果进行比较分析，研究了基于多特征空间自动提取方法的特点。第四章总结本文的研究内容，分析本文的创新之处，得出本文的研究结论。更多还原

【Abstract】 Remotely sensed imagery is the most immediate and abundant carrier of integrated information for ground target, and extracting terrain and thematic information from it automatically is an impending and complicated problem for remote sensing and GIS. Along with the quick development of remotely sensed obtaining technology in outer space, which provides more and more data, study personnel can get little effective information synchronously yet. Automatic extraction of thematic information from remotely sensed data has become bottleneck in Remotely sensed information production flow. Extraction method based on spectral character and single target figure character has been developed many years, but interpretation impact is short of anticipation desire despite of getting some evolution.This paper studied Remotely sensed information expression of object manifold recognition characters, analyzed allude relation between Remotely sensed imagery and target multi-characters space, put forward the study method of automatic Remotely sensed thematic information extraction based on multi-characters space. According to the evaluation for corn thematic information supervised classification result, with the analysis of influencing factors to corn spatial distributing, this paper brought forward automatic extraction model of corn thematic information based on multi-characters space, built the relevant expert knowledge storeroom and finished automatic classification of TM imagery thematic information using it. Comparing this classification result with the manpower visual interpretation, it was found that the former could improve efficiency of thematic information extraction, at the same time, this method was significant theoretically and also an innovation in automatic interpretation of remotely sensed imagery.This paper was divided into four chapters as follows: The first chapter summarized the present study and development direction about automatic interpretation of remote sensed imagery, and then explained the main study content, significance, aim and study method in this paper. The second chapter expatiated expression of manifold Remotely sensed information about ground targets, brought forward a method of thematic information extraction based on multi-characters space and constructed automatic extraction model of Remotely sensed thematic information based on multi-characters space. The third chapter took corn infield as researchobject, carried out thematic information extraction automatically by setting up corn extraction model, compared it with the extraction result of corn supervised classification, generalized the characters of automatic extraction method. The fourth chapter summarized study content of this paper, its innovation and stated the conclusion of this paper.更多还原