This paper proposed a method to identify the mechanical damage of fruit based on the combined hyper-spectrometer and electronic nose. We used hyper-spectrometer and electronic nose on no damage guava, lightlevel mechanical damage guava and heavylevel mechanical damage guava samples, respectively. After extracting the feature information, the principal component analysis (PCA), linear discriminant analysis (LDA), Euclidean distance (ED) analysis and fuzzy Cmean cluster were used to compare the classification effect of three identification methods (hyperspectral identification, electronic nose identification, combined hyperspectrometer and electronic nose identification) for guava mechanical damage. The results of PCA and LDA show that the hyper-spectrometer and electronic nose can identify the mechanical damage of guava, but both of the single method cannot identify the mechanical damage level of guava. When using the method of combined hyper-spectrometer and electronic nose identification, LDA result shows that it can classify the mechanical damage level of guava effectively. The identification effect of LDA for guava mechanical damage was better than that of PCA. According to PCA, LDA and ED results, we can also infer that the multisource information fusion can not only gain more sample information which was useful for improving classification effect, but also keep the maximum distance of each group as large as possible. According to fuzzy C-mean cluster results, the identification accuracy of guava mechanical damage based on hyperspectral identification, electronic nose identification and combined hyper-spectrometer and electronic nose identification were 89.74%, 82.05% and 97.44%, respectively. This paper proved the feasibility of using multi-source information fusion to improve the identification effect of fruit mechanical damage.