THE SYSTEM OF INTELLIGENT FRUIT QUALITY CONTROL IN INDUSTRIAL FRUIT AND VEGETABLE PRODUCTION

Industrial cultivation of fruit and vegetable products in open or protected ground is an integral strategy in achieving a multiple increase in production volumes and ensuring the country’s food security. The distribution of industrial volumes of harvested crops by commodity varieties and the gradation of fruits according to established quality indicators, including the maturity, is impossible without replacing the traditional visual assessment of this indicator with modern digital intelligent systems that reduce costs and increase production efficiency. (Research purpose) The research purpose is proposing the structure of the system of intellectual control of the ripeness of fruits and vegetables based on technical vision to minimize the set of work operations of preparation for eating fruits. (Materials and methods) Proposed a control system structure representing a distributed architecture with a key element in the form of a convolutional neural network implemented in Python using TensorFlow, NumPy, Keras and Matplotlib libraries. A sample of 1500 images of fruits (for example, tomatoes and others) distributed over three levels of maturity was used to train the network. The experimentally optimal architecture of the convolutional neural network and hyperparameters were determined: the size of the convolution kernel and optimization algorithms. (Results and discussion) It was shown that the size of the convolution kernel and the optimization algorithm critically affect the adequacy of the model. It was found that the 3×3 core and the Adagrad optimizer provide the best balance between local and global features and unambiguous classification. It was established that the developed model has the necessary generalizing ability and can be successfully used to control the degree of fruit maturity. (Conclusions) The structure of the system of intellectual control of fruit maturity based on technical vision, which ensures the efficiency of industrial fruit and vegetable production, including greenhouse production, was substantiated. It was stated that the developed system is conveniently integrated into the existing processes of agricultural enterprises to optimize harvest planning and improve their economic performance.