Maize quality prediction method based on the PIC-former model

Maize is one of the three major cereal crops in our country. During storage, maize quality can deteriorate. The speed of this deterioration is influenced by initial moisture content and storage conditions. It's essential to assess quality indicators during maize storage to determine the maize's state and better adjust storage strategies. This study considers time, initial moisture content, temperature, and humidity as independent variables, while multiple quality indicators of maize serve as dependent variables. We developed the PIC-former model based on the Transformer model to predict changes in maize quality. Additionally, the K-Means++ algorithm evaluates the quality during storage by comparing predicted values with actual values. We conducted experiments using maize storage data from Pinggu District in Beijing, and the results showed that the PIC-former model achieved the smallest prediction error. Therefore, this paper can provide technical support for maize storage and offer effective reference for warehouse personnel to adjust storage conditions.