Objective To effectively solve the problems of complex operation, long time consumption, insufficient accuracy, and difficulty in balancing speed and accuracy in existing quality inspection methods for refrigerated fish.Methods On the basis of a food quality inspection system that integrates multiple detection technologies, a rapid quality inspection method for refrigerated fish that combines hyperspectral data and machine vision is constructed. The hyperspectral imaging technology is employed to collect spectral data of refrigerated fish, and the data are preprocessed and input into an improved partial least squares regression model to achieve accurate determination of total volatile basic nitrogen content, and then the internal quality is graded based on the content. The machine vision technology is adopted to collect overall appearance images of refrigerated fish, and the images are input into an improved YOLOv13 model to achieve the grading of appearance quality. Internal inspection and external inspection are combined to comprehensively evaluate the quality of fish. Through comparative experiments, the superiority of the proposed method compared with existing methods is verified.Results The overall quality grading accuracy of the proposed fusion inspection method reaches 98.7%, and the improved PLSR model has a coefficient of determination being 0.970 for the total volatile basic nitrogen content, with a root mean square error of 0.15 mg/100 g. The improved YOLOv13 model has an appearance quality grading accuracy of 99.0% and inference time <20 ms/frame.Conclusion The proposed fusion method effectively integrates the internal physicochemical advantages of hyperspectral imaging with the rapid appearance detection capabilities of machine vision, significantly enhancing inspection accuracy and speed and meeting the needs of rapid batch quality inspection in processing stages.