Research On The Method Of Face Detection Based On Skin Color Model And Facial Landmarks Location

Face detection and landmarks location play an important role in the field of computer vision,and they are also very extensive in application of life.Face detection is to detect whether there is a face in the input image,if it exists,then identifies the location of the face area.Landmarks location is based on the face area detection,it is to find the key points of the face more accurately.Face detection is the first step of the whole face detection and landmarks location system,a precise face area positioning will make the follow-up landmarks location get higher reliability.In this paper,the skin color model and the face classifier network trained by the deep learning are combined for detecting the face area,on the basis of face detection,the regression network is used to return the part area with some inaccurate detection results to obtain more accurate face detection and positioning,next,the landmarks are located according to the corrected face region.The facial landmarks is based on the random forest method,and a global constraint model is established for the facial landmarks,then use the cascade regression structure to obtain the exact position of the facial landmarks step by step.The main contents of this paper are as follows:1?The traditional face detection method based on skin color model is usually applied to face detection task with Adaboost cascade classifier.However,when Adaboost is used for face detection,it is necessary to design and extract a large number of features from the human face in training and detection,and many weak classifiers are combined to form a strong classifier.This process requires manual design and extraction of a large number of features.In this paper,the depth learning method is used to design and train a human face classifier network,which can be combined with the skin color model to detect the human faces in complex scenes.2?For the case that there may be some areas of the face detection accuracy is not high in the previous step of face detection,we design the regression network to make this part of the detection area get the more accurate detection and positioning.3?The accurate face detection plays an important role in improving the speed and accuracy of the landmarks location,because we will give the landmarks the initial value according to the detected face regions.In this paper,the random forest model is trained to filter the important features of the landmarks,and establish the global constraint model ofthe facial landmarks,the least squares method is used to optimize the global model parameters.Finally,the exact position of the face feature points is obtained step by step using the cascade regression structure.The experimental results show that the improved face detection and facial landmarks location system can effectively improve the performance of face detection and facial landmarks location in complex environment.Besides ensuring the robustness and high positioning accuracy,it also has a near real-time high detection and positioning speed.