The invention discloses a face key point detection method and system based on local principal component analysis. The method includes steps: S1, collecting a large amount of face image sample data, and marking the face key points; S2, dividing the face key points Form a plurality of local key points, adopt principal component analysis to process each local key point respectively, obtain the principal component feature of each local key point; S3, calculate the combination of each key point of each piece of face image under the described principal component feature Coefficient; S4, build regression model, train model by described combination coefficient, generate combination coefficient regression model; S5, input the combination coefficient regression model of face image to be detected, predict and obtain described its combination coefficient; S6, based on prediction The obtained combination coefficients and the principal component features restore the key points of the face. The invention performs local principal component analysis on key points, predicts local principal component coefficients, reduces the complexity of directly performing principal component analysis on all key points, and improves regression modeling precision.