Scholarlink

In modern aviation, land and sea transport systems, computer vision systems and algorithms are widely used for navigation, safety, diagnostics, robotics, artificial intelligence in control and medical applications, including those based on image processing in various specific wavelength ranges (spectrozonal photography), mainly obtained using expensive multispectral and hyperspectral multichannel cameras. The purpose of this work is to investigate the possibility of evaluating the color spectral characteristics of an object (reconstruction of the shape of the spectrum) using data from three channels of a conventional RGB sensor. An algorithm for solving the pseudo-inverse problem using multidimensional optimization with constraints for the error of approximation of synthesized RGB data to the measured data is proposed. The results of spectrum reconstruction on a group of test examples are presented, which showed a fairly good convergence of this method for simple spectral forms.

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

Reconstruction the Shape of the Spectrum of Incident Radiation from Integral Sensors Data in RGB Range: Possibility Estimation and Uncertainty Reduction

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Alexander Grakovski,
  • Igor Radchenko

摘要

In modern aviation, land and sea transport systems, computer vision systems and algorithms are widely used for navigation, safety, diagnostics, robotics, artificial intelligence in control and medical applications, including those based on image processing in various specific wavelength ranges (spectrozonal photography), mainly obtained using expensive multispectral and hyperspectral multichannel cameras. The purpose of this work is to investigate the possibility of evaluating the color spectral characteristics of an object (reconstruction of the shape of the spectrum) using data from three channels of a conventional RGB sensor. An algorithm for solving the pseudo-inverse problem using multidimensional optimization with constraints for the error of approximation of synthesized RGB data to the measured data is proposed. The results of spectrum reconstruction on a group of test examples are presented, which showed a fairly good convergence of this method for simple spectral forms.