Wheat freshness recognition leveraging Gramian angular field and attention-augmented resnet

Abstract

During storage, wheat kernels undergo complex biochemical changes that affect their quality. Therefore, accurate and rapid measure of wheat freshness has immense economic and societal value. Using biophotonics and deep learning, this article explores the intricate relationship between wheat's ultra-weak bioluminescence signatures and its freshness. First, we select an advanced biophotonic system to capture time-varying bioluminescence data from kernels, which is then transformed into two-dimensional image styles employing the innovative Gramian angular field (GAF) method. Second, the image data serve as input to our proposed GAF-ResNet-GCT network architecture, which is specifically designed for wheat freshness classification and discrimination. The results underscore the effectiveness of our approach, demonstrating the model's remarkable ability to swiftly and precisely identify freshness with accuracy and robustness. The findings presented herein offer a groundbreaking scientific methodology for rapid, non-destructive wheat freshness detection, thereby advancing the application of biophotonics technology within the agricultural sector.

Keywords

• wheat kernels
• biophotonics technology
• gaf
• freshness
• stored wheat