Open AccessArtificial neural network technologies as a tool to histological preparation analysis
Author(s) -
Marina A. Nikitina,
И. М. Чернуха,
В. А. Пчелкина
Publication year - 2019
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/333/1/012087
Subject(s) - structuring , convolutional neural network , computer science , artificial neural network , sample (material) , identification (biology) , raw material , artificial intelligence , raw data , set (abstract data type) , quality (philosophy) , component (thermodynamics) , pattern recognition (psychology) , machine learning , biology , ecology , philosophy , chemistry , botany , physics , finance , chromatography , epistemology , economics , programming language , thermodynamics
Food fraud has become one of the great internal and external threats nowadays. For effective and objective detection of adulterated products, real-time control systems are used.Microstructural analysis is one of these methods. It allows rapid quality and composition assessment of food’s raw materials and the finished products. Human skills and expertise/involvement in interpreting the resultsof microstructural analysis is still absolutely necessary, despite the possibility this is subjective. Modern hardware/software systems for histological analysis with neural network technologies will exclude the subjectivity of that result interpretation. For convolutional neural network operating, the “learning with the teacher” model is used. With this training method, a set of data is prepared, whichacts as a series of observations, and for which the values of the input and output variables are indicated, such as: histological analysis data > conclusion about the sample composition > adulteration definition. The network learns to establish connections between them. This paper presents the classification parameters for vegetable food component identification as a part of meat raw materials and finished products. A unified database structure has been developedfor structuring and summarizing the main microstructural characteristics of histological sections for various types of meat raw materials. The production-rule systems of “ If … then … else ” are given, and vegetable protein components were chosen as an example. Based on these rules, training of convolutional neural network occurs.