ИНТЕЛЛЕКТУАЛЬНАЯ АВТОМАТИЗИРОВАННАЯ СИСТЕМА УПРАВЛЕНИЯ СОРТИРОВКОЙ ФРУКТОВ

Abstract

The development of effective technologies to reduce labor costs for the production, storage and processing of agricultural products is a difficult task. One of the promising directions of increasing the efficiency of technological processes in the agro-industrial complex is the full use of vision systems and artificial intelligence algorithms. Assessment of the quality of crop production is applied at various stages of production and allows detecting its damage, including during sorting. Due to the high productivity of production lines at the facilities of the agro-industrial complex, it seems effective to use intelligent information and control systems operating on the basis of machine learning algorithms. (Research purpose) The research purpose is proposing an approach to the creation of an intelligent automated control system for the fruit picking process using a vision system. (Materials and methods) Considered the structure of the information and control system of fruit sorting using a system of technical vision and machine learning. We implemented the test architecture of the neural network using the Keras library based on the Python language, the research was carried out in the google.collaborate service. A prepared data set based on images obtained on the production line was used to train the neural network. (Results and discussion) The structure of the convolutional neural network architecture was obtained, which allows us to adequately and with a sufficient accuracy solve the problem of classifying fruits into healthy and damaged ones in order to integrate it into an intelligent information and control sorting system. (Conclusions) Proposed an approach to improving the fruit sorting process, for example, for further processing into juice, using a technical vision system and machine learning. The structure of a convolutional neural network was developed to solve the problem of classifying apples by appearance, which became the basis for creating an intelligent information and control system for assessing the area of damage to the surface of fruits.