Tomato Detection and Classification by Color, Size, and Imperfections using Computer Vision

Abstract

Listen

Translate article

Tomatoes stand out in agribusiness, as they are one of the most consumed vegetables in the world. There is a growing search for healthier foods and tomatoes are an important source of vitamins A, C and B1. Its production, handling and transportation are complex due to its sensitivity to the environment, pest attack, road conditions, among others. Consumer markets are increasingly demanding, with its quality as a decisive factor for purchases. The price is also one of the primary factors for the purchase by the customer, both the prices and the quality of the tomato take into account the size, color and quantity of imperfections that the fruit presents. The classification by size nowadays is done using a mat with holes of small, medium and large sizes and as the tomato fits in the mouth whose size best adapts it will fall into boxes with tomatoes of its respective size, however, on these mats it is not possible to check neither the color nor the imperfections. Another form of classification is by electronic equipment that is capable of verifying the size and color, but they are imported machines and extremely expensive, in addition to not detecting imperfections. The most common form nowadays is the manual one, where a trained person separates the tomato according to its size, color and imperfections, however, this form of analysis makes the classification process slow and subject to subjectivity. In view of this scenario and taking into account the numerous applications of Computer Vision in agribusiness, this work proposes to create a tomato classifier by color, size and imperfection, using computer vision.