Automated Identification of Citrus Fruits Nutrients Through Non-destructive Analysis

Abstract

The development of digital technology has played a significant role in digitizing this world. This technology has enabled the storage, processing, and analysis of data using a cloud-based platform that benefits farmers. Image processing is a technique that involves performing specific tasks on an image to extract useful information. Citrus fruits such as mandarins, lemons, grapefruits, and oranges are the most widely grown fruits in the world. Citrus is a large plant cultivated primarily in the world's tropical regions due to its abundance of vitamin C, Total Soluble Solids (TSS), Titratable Acidity (TA), and pH. To find out these nutrients, we need to consult with a horticulturist.  The traditional approach is costly, complex, and time-consuming for a common farmer. The study aims to provide an efficient and cost-effective solution to achieve the same goal. This work trains an efficient deep learning-based system to process data more efficiently and precisely. A framework is developed to automate the system for classifying and detecting nutritional values from images of different citrus fruits. The system is trained using the Transform Learning Approach (TLA) and Single Shot Detection (SSD) V2 to process the custom dataset. Results obtained from the experiments show that the accuracy achieved with our proposed methodology approaches 97%. This non-destructive predictive analysis of citrus fruit will pave the path for prescriptive analytics to enhance the qualitative productivity of the fruit.