Abstract

Food quality assessment plays a crucial role in the modern food supply chain because it directly influences consumer safety, product value, and regulatory compliance. Traditional inspection procedures rely heavily on human expertise, which introduces subjectivity, inconsistency, and increased operational cost. The growing adoption of edge computing has created an opportunity for automated food quality evaluation systems that can operate with minimal computational resources. However, many existing deep learning approaches require substantial memory, processing power, and energy consumption, which limits deployment on edge devices. To address this challenge, this study proposes an innovative framework named Histogram-Guided Efficient Neural Assessment Network (HGENAN) for real-time food quality assessment. The proposed HGENAN method integrates histogram-based image characterization with an efficient deep neural architecture to extract discriminative quality features from food images. Initially, an adaptive histogram enhancement module improves visual contrast and highlights quality-related attributes. Subsequently, a histogram-guided feature extraction unit generates statistical descriptors that complement the learned representations from a lightweight convolutional neural network. A feature fusion strategy combines handcrafted and deep features, while an adaptive classification layer identifies food quality categories with reduced computational complexity. The proposed Histogram-Guided Efficient Neural Assessment Network (HGENAN) is evaluated using a food quality image dataset containing 12,000 samples from three quality categories. Experimental results demonstrate that HGENAN achieves an accuracy of 98.7%, precision of 98.5%, recall of 98.4%, and F1-score of 98.4%. The proposed framework reduces inference latency to 14.2 ms per image and decreases computational requirements by 36.8% compared with conventional deep learning approaches. The obtained results confirm that the integration of histogram-based statistical descriptors with a lightweight neural architecture provides an effective solution for accurate and real-time food quality assessment on edge devices.

Authors

Sinjan Kumar¹, Subodh Kumar²
Government Engineering College Vaishali, India¹, Katihar Engineering College, India²

Keywords

Food Quality Assessment, Edge Artificial Intelligence, Histogram Analysis, Lightweight Deep Neural Network, Computer Vision