Abstract: The effect on the ribbonfish freshness detection results was studied using diffuse reflection and diffuse transmission of NIR, and minced and complete ribbonfish.The diffuse reflection and diffuse transmission spectrum of minced ribbonfish and the diffuse reflection spectrum of complete ribbonfish were collected respectively using integrating sphere diffuse system in the range of 4000 ~ 10000cm- 1.The model parameters were optimized according to the cross- validation mean square error ( RMSECV) of training set and the correlation coefficient r.Quantitative detection models of total volatile basic nitrogen ( TVBN) , pH and total number of colonies were established respectively by partial least squares ( PLS) .Predictive ability of the models were verified by prediction mean square error ( RMSEP) , correlation coefficient r and paired t- test.Results showed that prediction performance of the models for the determination of the freshness of ribbonfish was good.But near infrared diffuse reflection compared with diffuse transmission, the prediction ability of the former was higher than the latter.The model's prediction ability of minced ribbonfish was higher than that of the complete ribbonfish when using diffuse reflection spectrum.Therefore, diffuse reflection spectrum of minced ribbonfish was appropriate when testing ribbonfish freshness.