Abstract
This study addresses the challenges of large sample size dependency and sample imbalance in traditional pork color scoring models. We propose a rapid method for constructing an accurate color scoring model using six standard color board images and compare its performance with traditional models based on 525 real pork samples from seven pig herds. The results show that the classification accuracy of the CS_1 models, after intercept calibration with mixed herd images, is comparable to traditional models. Specifically, accuracies for CS_1_L, CS_1_La*, and CS_1_Lab models within a ± 0.50 scale are 91.43%, 95.62%, and 94.10%, respectively. Calibration using individual herd images significantly improves accuracy, with CS_1_L, CS_1_La*, and CS_1_Lab* models achieving accuracies of 93.75%, 95.90%, and 96.10%, respectively. This method offers advantages such as small sample sizes and rapid intercept calibration, providing a new approach for objective pork color assessment.