Scholarlink

<p>The article addresses the issue of automating the interpretation of remote sensing data on natural features. The existing methods of structural analysis of spectral data are shown to rely on expert parameter evaluation of the used algorithms. In order to automate the interpretation of remote sensing data, a&#xa0;modified structural analysis method using the components of a&#xa0;correlation coefficient for a&#xa0;pair of spectral features was developed. Each component is defined by the product of its constituents in the form of normalized spectral features. According to the signs of the components comprising the correlation coefficient (positive, negative, and alternating), four classes are identified. The obtained data is used to establish a&#xa0;decision rule for assessing whether a&#xa0;test situation in the space of a&#xa0;spectral feature pair belongs to one of the identified classes. On the example of identifying forest areas damaged by the Siberian silk moth, the authors compared the results of applying the proposed method and conventional methods for the decomposition of remote sensing data, which use NDVI (normalized difference vegetation index) and GNDVI (green normalized difference vegetation index). In order to characterize the classes identified via the modified method and determine the threshold values of NDVI and GNDVI, kernel density estimates were used. The article presents a&#xa0;procedure for optimizing the kernel density estimate that relies on selecting blur coefficients in kernel functions based on the maximum likelihood function. The modified structural analysis method helps to circumvent the problem of determining threshold values for difference vegetation indices.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Modified method of structural analysis of remote sensing data

  • 期刊论文

  • 发表日期：

    2026年5月6日

查看全文
  • Aleksandr V. Lapko,
  • Vasiliy A. Lapko,
  • Sergey T. Im,
  • Yuri P. Yuronen

摘要

The article addresses the issue of automating the interpretation of remote sensing data on natural features. The existing methods of structural analysis of spectral data are shown to rely on expert parameter evaluation of the used algorithms. In order to automate the interpretation of remote sensing data, a modified structural analysis method using the components of a correlation coefficient for a pair of spectral features was developed. Each component is defined by the product of its constituents in the form of normalized spectral features. According to the signs of the components comprising the correlation coefficient (positive, negative, and alternating), four classes are identified. The obtained data is used to establish a decision rule for assessing whether a test situation in the space of a spectral feature pair belongs to one of the identified classes. On the example of identifying forest areas damaged by the Siberian silk moth, the authors compared the results of applying the proposed method and conventional methods for the decomposition of remote sensing data, which use NDVI (normalized difference vegetation index) and GNDVI (green normalized difference vegetation index). In order to characterize the classes identified via the modified method and determine the threshold values of NDVI and GNDVI, kernel density estimates were used. The article presents a procedure for optimizing the kernel density estimate that relies on selecting blur coefficients in kernel functions based on the maximum likelihood function. The modified structural analysis method helps to circumvent the problem of determining threshold values for difference vegetation indices.