The modeling research of wheat classification based on NIR and RBF neural network

Since the traditional detection method of wheat quality was tedious and time-consuming, near infrared reflectance spectroscopy (NIRS) combined with RBF artificial neural network was used to classification and detection wheat quality non-destructively and quickly in this paper. The ware point of samples obtained by the NIRS is too many, resulting in the structure of RBF neural network is too complex, so we used the algorithm of radial basis function (PSO) to optimize RBF neural network, and made some improvement measures against the shortcoming of premature convergence and the set of inertia weight was too mechanical of PSO algorithm. The experimental analysis showed that the accuracy of model can reached 98%, which could satisfy the need of non-destructive and real-time detection of wheat in modern agriculture.