Abstract: Infrared spectroscopy was used to carry out non-destructive testing of walnut to realize identification of seven walnut cultivars，such as Qingxiang，Wen185，Xiangling，Xinxin No.2，Zhipi，Yangbi and Daifeng，to provide new ideas for walnut cultivars identification in the future.A method was proposed based on near-infrared and far-infrared spectral information fusion.Far-infrared spectroscopic technology was applied to walnut identification for the first time and fused near-infrared spectroscopy data.Combined with principal component analysis（PCA）and uninformative variables elimination and successive projections algorithm（UVE-SPA），characteristic wavelength was selected.Then random forest，K-nearest neighbor and support vector machine classification model was established.In addition to random forest model，recognition accuracy of other models could reach 100%，which not only reduced complexity of model but also greatly improved accuracy and robustness of identification.Test results showed that，data fusion based on effective information of near-infrared spectrum and far-infrared spectrum could improve deficiency of single spectrum technology in recognition rate，that was，data fusion of two bands could better reflect differences between varieties.It provided a new idea for efficient，non-destructive and accurate identification of walnut cultivars，and also a reference for identification of other substances.