New AI System Quickly Assesses Dried Squid Quality Without Damage
TOKYO – A new artificial intelligence system promises to rapidly and non-destructively assess the quality of dried squid,a crucial food product in many Asian countries. Researchers have developed a method combining hyperspectral imaging – analyzing light across a wide spectrum – with a specialized deep learning model called a 1D-KAN-CNN (one-dimensional Kolmogorov-Arnold network convolutional neural network). The technology allows for the measurement of fat content, protein levels, and total volatile basic nitrogen – key indicators of freshness and quality – without physically altering the product.
Demand for dried squid is significant globally, and current quality control methods are often time-consuming, expensive, or require destroying a sample. This new system offers a swift, noninvasive choice for the food industry, possibly reducing waste and ensuring consistent product quality. The research, detailed in a recent paper, analyzed 93 dried squid samples using VIS-NIR (400-1000 nm) hyperspectral reflectance imaging. Critical wavelengths were identified using a combination of competitive adaptive reweighted sampling, principal component analysis, and the successive projections algorithm, feeding data into the 1D-KAN-CNN model for accurate assessment. This breakthrough could streamline quality control processes for producers and distributors, ultimately benefiting consumers.
