Research on face feature point detection algorithm under computer vision

The face feature point detection algorithm has a very broad application prospect in the field of face recognition. In this paper, from the perspective of computer vision, the specific connotation of face feature point detection and face shape indexing with pixel coordinates is elaborated. The overall framework of cascade regression algorithm is built through training and testing, and the face pose changes are extracted based on pose indexing features and weak invariance to ensure that the algorithm can realize real-time face feature point detection. Train all levels of cascade regressors using data distribution statistics to allow the cascade regression framework to complete incremental learning and restore the position of real shape markers. Compare the cascade regression algorithm with existing face feature point detection methods, and verify that the cascade regression algorithm has high detection accuracy and fast detection speed in face feature point detection by pupil localization test and detection speed test. The detection rate of the cascade regression algorithm in the test is more than 90%, the pupil detection accuracy can be controlled at about 3 pixels, and the detection speed is about 87FPS, which is able to quickly and accurately recognize the face feature points, and it has practical and broad application prospects in real life.