The work aims to establish the discriminant models of Fisher, K nearest neighbor (KNN) and partial least-square discriminant analysis (PLS-DA) for the pressed apples based on ultravioletradiation/visible spectroscopy technology combined with pattern recognition algorithm. With pressed apple and intact apple as the study object, the optical fiber spectrum system was used to acquire the spectral reflectance of these two kinds of apple. The effect of different spectral pretreatment methods (Second Derivation (SD), Standard Normal Variation (SNV) and Multi-Scatter Calibration (MSC)) on the recognition effect of each model was comprehensively compared, and the principal component analysis (PCA) was applied for the dimensionality reduction of the pretreated spectral data, and the characteristic spectrum that could reflect the pressed apple was extracted. The results showed that the first 7 principal components (P1—P7) with cumulative contribution rate of 99% were selected as the characteristic spectral data by the principal component analysis, and the dimensionality reduction of the spectral data was well realized. The preprocessing effect of Second Derivation on spectral reflectivity was the best, and three models could all meet the practical requirements, especially both SD+Fisher and SD+PLS-DA models had the optimal recognition performance with a total correct recognition rate of 100% for the samples of calibration set and prediction set. The research results are conductive to the fast recognition of pressed apples.