DEVELOPMENT OF METHODS FOR MULTISPECTRAL DIFFERENTIATION OF NATURAL AND SYNTHETIC MATERIALS BASED ON SPECTRAL CHARACTERISTICS OF DIFFUSE REFLECTION DITIONS
An analytical review of the existing problems for organic (i. e. plant leaves) and synthetic functional composite materials (FCM) differentiation by existing ground-based remote terrain sensing methods, as well as methods for implementing discrimination algorithms for organic and synthetic FCM based on multispectral and hyperspectral images obtained by photometric method has been carried out. As a result, it has been established that the development of algorithms for the differentiation of organic and synthetic FCMs operating in the spectral range of wavelengths from 300 to 900 nm is an actual task in the development of ground-based methods for remote terrain sensing. Purpose – to determine the possibility of differentiating organic and synthetic PCM by means of multispectral terrain sensing in the spectral range of wavelengths from 300 to 900 nm. The article used diffuse reflectance spectroscopy, localized polynomial regression (LOESS) and residual sum of squares (RSS) methods for processing and analyzing diffuse reflectance spectra, backpropagation method, Adam optimizer for materials differentiation algorithms and Accuracy metric. Authors researched the spectral reflective characteristics of some plants and synthetic functional composite materials intended for their imitation. Two multispectral algorithms for the differentiation of FCM materials have been developed, and a comparative analysis of the correctness of the algorithm’s «accuracy» revealed that the most effective parameter for assessing the spectral characteristics of materials is the integral parameter G. It is advisable to apply the obtained results as an auxiliary criterion of existing ground-based methods of remote terrain sensing, carried out to discriminate between organic and synthetic functional composite materials by means of their diffuse reflection spectral characteristics.
Authors: Sh. Z. Lovpache, Yu. V. Mamelin, S. G. Sinitsa, V. Yu. Buzko, A. S. Mamelina
Keywords: Diffuse reflection spectroscopy, photogrammetry, multispectral imaging, bandpass filters (FSF), material differentiation, vegetative indices, neural networks
View full article