Identification of peach kernel and almond kernel medicinal materials based on near-infrared spectroscopy and 1D-CNN

Objective To address the limitations of traditional identification methods in distinguishing between peach kernel and almond kernel, our research team has developed a rapid and precise discrimination approach. Methods Near-infrared spectroscopy (NIRS) was employed in combination with neural networks to establish a discrimination model. A self-developed miniaturized NIR spectral acquisition device was utilized to collect data, which underwent preprocessing including smoothing and first-derivative processing. Principal Component Analysis (PCA) was applied for dimensionality reduction, followed by modeling using a 1D Convolutional Neural Network (1D-CNN) to discriminate between peach kernel and almond kernel. Results The NIR spectra of Peach kernels and almond kernels exhibit similar absorption peak shapes. However, the 1D-CNN discrimination model effectively differentiated between the two species, demonstrating strong predictive performance and high fitting capability. Conclusion This study presented a rapid, low-cost, and accurate method based on NIRS for discriminating between peach kernel and almond kernel.