Research On Face Detection And Facical Point Detection Based On Convolutional Neural Network

With the rapid development of digital image technology and deep learning theory,it has become a hot topic to use deep learning algorithm for digital image acquisition.As an effective feature,face has been paid much attention.This paper focuses on the use of deep convolutional neural network model in the field of face detection and facial point detection.It proves more efficient if we expand training samples and construct cascade structure.The main content of this paper is as follows:(1)Face detection algorithm.Firstly,this paper introduces a face detection algorithm based on convolutional neural network(CNN).In order to improve the robustness of the algorithm,this paper proposes a training sample expansion algorithm,which makes the algorithm a better detection effect on face gesture,occlusion and illumination.In order to remove false faces in the image Pyramid model,we introduce non maximum suppression algorithm in solving this problem,and successfully reduces face detection rate.Then we focus on the a method of convolutional neural network cascade for face detection,and propose the use of Relu activation function and Dropout regularization.It not only improves the convergence speed of the network,but improves the generalization ability.(2)Facial point detection algorithm.Firstly,facial point detection algorithm based on convolutional neural network is studies.In order to further improve the detection effect,a from coarse to fine double layer cascade convolutional neural network is proposed to solve the problem.In this model,firstly,we propose a sample expansion strategy to enhance the robustness of the system;secondly,the gradient feature is proposed in the first layer of the cascade structure to construct a parallel network model of the pixel domain and gradient domain.Then the two detection results are weighted with best fusion ratio.The experimental results show that the improved algorithm has a lower calibration error,and it has a better effect on the face,facial expression,skin color,occlusion and illumination.(3)Design and implementation of face detection and feature point detection system.Firstly,we design the system from system performance and software architecture.Secondly,we show the implement details of two core modules,including face detection and facial point detection.At last,we use MATLAB to achieve core code,and design GUI software interface for interactive display.