Abstract: Principal Component Analysis (PCA) algorithm can simplify high-dimensional problems into low-dimensional problems. It is simple and fast, and the principal components are orthogonal to each other, which can eliminate the influence of the original data components. The face recognition technology based on PCA algorithm can remove noise caused by light, posture, and occlusion to some extent. The Support Vector Machine (SVM) method using kernel function can solve the nonlinear problem and has perfect classification e ect. In this paper, combined with the PCA and SVM methods, dimension reduction and feature extraction are performed on the untrained images, and then the features are input into the SVM using the Gaussian kernel function for training. The performance of the SVM classifier is verified using 10-fold cross validation method. This method is suitable for scenes with high requirement for recognition speed, such as unmanned patrol car in industrial park.