Research on the Application of Deep Learning Framework Enhanced with GA-PIO Hybrid Algorithm in Intelligent Rice Variety Classification

Rice is one of the main staple foods worldwide, and accurate variety classification is a critical aspect in its production and distribution. This study established a dual-layer optimization framework that integrates the GA-PIO fusion algorithm and deep learning technology, aiming to achieve precise classification of rice varieties. This framework encompasses three core stages: feature extraction, optimization of the external network structure, and internal parameter tuning, utilizing Bayesian methods and the GA-PIO fusion algorithm for layer-by-layer network optimization. Moreover, a dynamic variance threshold adjustment adaptive method has been introduced within the GA-PIO, enhancing the diversity of the population and avoiding premature convergence to local optimums. Through experiments, the BP model achieved an average classification accuracy rate of 93.10%.