Apple Brand Texture Classification Using Neural Network Model.

This paper describes a texture measurement method for apples and a neural network model to infer the texture classification in apple brands. The features such as load and sound measured by the texture inspection equipment are input to the neural network model. The model identifies the corresponding apple brand. The authors had measured the apple texture effortful manner in our previous paper. For instance, we had to hollow out the sample of apple flesh using a stainless-steel pipe, and then we cut the flesh in the same size for texture inspection. This paper describes an improved method, where the inspector will slice the entire apple into the same thickness to examine the texture of flesh. We experimented with the proposed method to obtain the load and sound signals for three brands of apples: Sun-Fuji, Orin, and Shinano-Gold. The load and sound features are input to the neural network model to classify three apple specimens. Even though the features were complicated to distinguish the specimens, the neural network model could identify the corresponding specimens. This paper shows the validation result of the proposed method and the neural network model and describes future works.