Freshness Grading of Agricultural Products Using Artificial Intelligence

Introduction

In the rapidly developing world of the twenty-first century, food delivery and freshness control have become a major issue because of the fast increase in the population. Thanks to modern technologies, people started to spend less time grocery shopping and take advantage of the comfort brought by online shopping. Among the online orders, fruit and vegetable orders constitute a significant percentage. However, in online fruit and vegetable shopping, the end consumers cannot apply their choices and preferences when selecting goods, and customer satisfaction is highly dependent on products that are chosen by the supplier. Therefore, to provide customers better shopping experience, online groceries should determine the freshness and quality of the product in a systematic, consistent, and objective way and price them accordingly. To achieve this, it is necessary to evaluate the freshness level with measurement methods, and physical properties, in other words, the analysis should be quantitative. Using quantitative analysis, end consumers will be able to do fruit and vegetable grocery shopping online with the freshness level they prefer, leading to more comfortable and healthy shopping experiences. In this way, customer satisfaction is expected to increase. Furthermore, by detecting the fruits and vegetables that are soon to be wasted, the grocery can rearrange their price and prevent them to be wasted in the store. This not only is advantageous for the grocery but also helps to reduce wastage.

Among the studies about measuring the freshness of fruits and vegetables in the literature, some studies damage and waste agricultural goods during the measurement process. On the other hand, studies that do not damage or waste the goods are insufficient when applied alone. Therefore, analyses and test methods need to be applied and brought together harmoniously and systematically to obtain a useful tool. This research aims to bring these analysis methods together with the use of deep learning technologies and contribute to the existing literature with an innovative approach. In this chapter, the authors consider the analysis of the color, shape, texture, and size properties of agricultural goods are considered in the deep learning model and provide a method to measure the freshness of agricultural products in an objective way.

The freshness levels of agricultural products can be measured using chemical and physical test methods. For instance, for freshness detection of asparagus, the concentration of asparagine amino acid in the tip is a significant indicator while the fluorescence of chlorophyll is important for the detection of the freshness of broccoli. Nevertheless, chemical freshness analysis methods are not practical and not suitable in all conditions. Therefore, chemical tests are not commonly used among groceries and consumers. Furthermore, the freshness perception of consumers includes various quantitative and qualitative factors, such as size, shape, color, firmness, smell, and texture (Péneau, 2005). These factors cannot be generalized to cover all agricultural products at one time. To the best of the authors’ knowledge, there does not exist a comprehensive method that uses all the listed features to estimate the freshness levels of goods in the literature. On the other hand, analysis by physical inspection may damage the items and make them inconsumable. As a result, physical methods are not widely used in the freshness analysis of agricultural products.