GRAPE QUALITY PREDICTION IN PRE - POST HARVESTING WITH IMPLEMENTATION OF FUSION DEEP LEARNING

Abstract

The act of tilling and nurturing the land, engaging in the growth and harvest of various crops, and managing the rearing of domesticated animals is commonly denoted as the agricultural enterprise of farming. The significance of agriculture is indispensable in enhancing a nation's economic growth and development. Fruit is considered a type of perishable agricultural commodity. The Vitis genus, consisting of approximately 60 to 80 species of climbing plants within the Vitaceae family, is indigenous to the northern temperate region. This genus encompasses cultivars that can be consumed as fresh produce, sun-dried for the production of raisins, or extracted for the manufacturing of grape juice or wine. The utilization of Artificial Intelligence (AI) has become prevalent within the realm of precision agriculture as a means of assessing the food quality. This notion is particularly pertinent when evaluating crops during different stages of harvest and postharvest. It is imperative that fruits are harvested at an appropriate stage of maturity to ensure optimal quality and enhanced storability. Convolutional neural networks (CNNs), which belong to the category of feedforward artificial neural networks, have demonstrated efficacious implementation in the field of agriculture for the purpose of image segmentation and object classification. The application of these techniques possesses the potential to facilitate the identification of objects of significant relevance, such as fruits or leaves, in agricultural imagery. Consequently, it presents itself as a viable and efficacious approach towards addressing the inherent obstacles encountered in large-scale agricultural operations. The present investigation puts forth a novel grape quality forecast model utilizing machine learning methods and image processing techniques to ascertain the ripeness and size of grapes. The corpus was generated through the acquisition of photographic representations of diverse grape cultivars at varying growth phases, harvested from assorted vineyards. The utilization of convolutional neural networks (CNNs) in modelling has resulted in the attainment of remarkable accuracy levels in grape size and ripeness prediction. The investigation resulted in a noteworthy model that attained an optimal precision level of 100% in anticipate grape dimensions and maturity. Prospective research endeavours could centre on enhancing the performance of the model, augmenting the scope of the dataset, and creating an instantaneous prognostication system for the management of vineyards