Classification of oatmeal residue levels on bowls using fluorescence imaging: Toward an intelligent method for...

The ability of dishwashers to remove food soiling poses an important challenge for food safety. Cleaning performance index (CPI) serves as a pivotal metric for evaluating food residues remaining after dishwashing. Currently, CPI relies on subjective visual inspection. To improve objectivity, this study introduces a fluorescence imaging-based method. Rice bowls soiled with oat flakes-made porridge were selected as the target to be tested. Using oat-porridge-soiled rice bowls prepared under surveyed temperature-time conditions, excitation-emission matrix (EEM) analysis identified an optimal 365-nm excitation wavelength for fluorescence imaging. You Only Look Once (YOLO) and Focal Modulation-YOLO (FM-YOLO)-based recognition model showed that the Blue-channel after chromatic adaptation (CA_Bchannel) performed best. Compared to YOLO, FM-YOLO demonstrates superior performance across three key metrics: P, R, and mAP@0.5. A correction method was proposed to eliminate the influence of the bowl-shaped surface on the calculation of the soiled area, and based on this, a cleaning performance evaluation method was established using the soiling area obtained by FM-YOLO. The CA_Bchannel dataset outperforms the raw dataset in cleaning performance classification, particularly for classes 2 and 4. After area correction, its accuracy reaches 86.7%, with a 1.7% improvement without correction. The accuracy for the dishes with scores of 5, 4, 3, 2, 1, and 0 using the CA_Bchannel image was 100%, 100%, 85%, 65%, 70%, and 100%, respectively. The proposed method offers a more objective approach for detecting food soiling, promotes improvements in the cleaning performance of dishwashers, and thereby provides stronger safeguards for public safety.