ABSTRACT
Bananas are a very delicious and nutritious fruit. Every year there are a large number of bananas sold at home and abroad and loved by the people. It is necessary and meaningful to study the freshness of bananas. Understanding the process of change will provide important guidance. Manual identification of change processes is time-consuming, expensive, and requires experienced experts, which is usually limited. This chapter analyzed the feasibility of using transfer learning to identify the changing process of fresh banana and established the relationship between fruit quality and time. We use a pretrained GoogLeNet model to identify the process of banana change. The experiment was carried out at room temperature for 11 days to determine the change process. The result shows that the correct recognition rate of this method is 92%, which is higher than human level. In order to make the experimental model universal, we also used the model to detect the change process of strawberries. In order to expand the application of transferred learning image recognition, we will conduct more research on the change process of fruit and vegetable varieties in the future, so as to establish a complete fruit and vegetable freshness detection system.
