Pistachio Classification: A Synergistic Approach of 1DCNN and Handcrafted Features

Accurate classification of pistachio varieties is essential for the agricultural economy, given their distinct characteristics and commercial value. This study introduces a novel two-stage framework to address this challenge. First, we construct a comprehensive feature representation by leveraging the synergy of three complementary handcrafted descriptors: basic geometric and color features, Local Binary Patterns (LBP) for fine-scale micro-texture characterization, and the Gray Level Co-occurrence Matrix (GLCM) for statistical texture analysis. Second, we design a specialized one-dimensional Convolutional Neural Network (1DCNN) architecture tailored to effectively process the fused feature vector. Extensive experiments on a real-world dataset demonstrate that the proposed method achieves a classification accuracy of 99.53%, outperforming conventional machine learning baselines and other deep learning variants. These promising results highlight the method’s potential for intelligent agricultural processing and suggest its applicability to broader precision agriculture tasks.