Identification of rice varieties based on spectral-image fusion using hyperspectral imaging technology

To ensure consumers purchase high-quality rice, accurate identification of rice varieties has become particularly important. This paper conducts research based on machine learning algorithms from the perspectives of images, spectroscopy, and spectral-image fusion. Hyperspectral image data of six different types of rice were collected, and preprocessing of the Region of Interest (ROI) was performed using Savitzky-Golay (SG) smoothing and Multiplicative Scatter Correction (MSC). Texture information of the images was extracted using the Gray-Level Co-occurrence Matrix (GLCM), obtaining eight texture features. In the spectral-image fusion process, spectral and texture data were combined into a new matrix. Using spectral, texture, and fused spectral-image data as model inputs, classification models such as Support Vector Machine (SVM), Logistic Regression (LR), and K-Nearest Neighbors (KNN) were constructed to classify rice varieties, and these models were compared. The classification performance of the fused spectral-image data was superior to models using only spectral or texture data. The fusion of texture parameters and spectral features enhanced the effectiveness of model evaluation, with detection accuracy surpassing that of models relying solely on spectral data.The research results show that the accuracy of both SVM and LR classification models exceeded 90%, with the LR model performing the best, efficiently classifying rice varieties.