Research On Face Detection Technology And Implementation

Face detection is a smart technology which uses the computer to detect the position,posture and size of the face in an inputted picture,and it has been widely used in identity authentication,safety protection and human-computer interaction.Because human face is a non-rigid object in a complex environment,face detection has met many difficulties in the process of using and spreading.In order to find a universal detection algorithm,domestic and foreign scholars have done a lot of work,and they proposed many kinds of schemes.To solve this problem,this thesis analyzed and compared several main detection algorithms,and then we studied the skin color detection technology and the face detection technology based on Adaptive Boosting algorithm deeply,and the main work of our thesis is as follows:(1)Firstly,this thesis introduces the background and the research status of the face detection technology in detail,and summarizes the system classification of the face detection technology.Then the face detection technology based on skin color segmentation,the face detection technology based on AdaBoost algorithm and the face detection technology based on neural network are introduced for future work.(2)In order to improve the false detection rate of the AdaBoost algorithm,a face detection scheme based on skin color segmentation and DF-AdaBoost algorithm is proposed Aiming at the disadvantages of AdaBoost algorithm in classifier weight distribution,the DF-AdaBoost algorithm based on double factor decision uses the "quality" factor and "quantity" factor to assess the performance of weak classifiers jointly.In this algorithm,several weak classifiers which have better performance are combined into a strong classifier to realize the face detection.We conbine the DF-AdaBoost algorithm with skin color detection technology and implement it in programming.We analyze the detection rate,false alarm rate,detection speed and robustness of the proposed scheme with the experimental result.It shows that the detection rate and false detection rate of the face detection scheme based on skin color segmentation and DF-AdaBoost algoritlim are 92% and 5.3% respectively,and our scheme has higher detection rate and lower false detection rate in the detection of the frontal faces,and the detection speed is fast,but the robustness still has room to improve.(3)Because the BP neural network has slow detection speed to detect a face,we propose a novel face detection scheme based on sparse features and BP neural network.We implement this scheme in programming and then analyze the detection rate,false alann rate,detection speed and robustness of the proposed scheme.The experimental result shows that the detection rate and false detection rate of this face detection scheme based on sparse features and BP neural network are 91% and 7.6% respectively,and our scheme has higher detection rate,lower false detection rate and stronger robustness in the detection of the multi-pose faces,and comparing with the BP neural network,the detection speed of this face detection scheme is faster.