In order to solve the problem of fast detection of adulteration of mutton and pork, the spectral collection of adulterated mutton was carried out by using a multi-spectral instrument, and the reflectivity of samples at the band of 350~1 100 nm was obtained. For data preprocessing, Particle Swarm Optimization (PSO) was used to optimize the Least Squares Support Vector Machine (LSSVM), and a Least Squares Support Vector Machine (PSO-LSSVM) model based on Particle Swarm Optimization was established, compared with Partial Least Squares(PLS), Back Propagation Neural Network (BPNN) and LSSVM models. The result showed that PSO algorithm could effectively optimize LSSVM model,and the decision coefficient and root mean square error of prediction was 0.920 4 and 0.089 2. Furthermore, Random frog (RF), Uninformative Variable Elimination (UVE) and Competitive Adaptive Reweighed Sampling (CARS) were used to extract the characteristic wavelengths and establishing the model of PLS. The results showed that the UVE-PLS model’s decision coefficient and the root mean square error of prediction set were 0.996 7 and 0.016 2, and UVE was better than other feature wavelengths extraction methods.