Abstract: Seed is a fundamental resource in agricultural production.Seed vigor has a close relation with germination and growth of the plant.Traditional methods of detecting seed vigor are timeconsuming,expensive and difficult to realize automation and large-scale application.Therefore,with the improvement of agricultural modernization of our country,there is a crucial need for a fast,accurate and non-destructive seed detection technology.This paper reviews non-destructive seed detection researches with near infrared spectroscopy,hyperspectral imaging,Raman spectroscopy and X-ray.Partial least square regression is still the most popular model to processing spectral data and some new methods,such as machine learning and deep learning,are introduced to establish spectral models.In terms of experiment designment,a two-category pattern to distinguish viable and nonviable seeds is the main trends in this field and threshold is closely related to the discrimination result.Quantitative models have a poor performance in seed vigor prediction and the slow change in seed vigor need to be further researched.For seeds of some species,characteristic wavelengths which reflect the ability of germination have been found with feature selection methods.As for apparatus,there is a lack of specially-designed device for spectra collection of single seed sample.The advantages of spectral and imaging technology in detecting seed vigor includes high speed,low cost and non-invasiveness,and they are suitable for automation and intelligence in agricultural production.