Researches On Face Detection And Recognition Based On Haar-Like T Features

Face detection and recognition is one of the most common biometric identification technology which is widely used. In recent years it has become more and more active research topic in pattern recognition, image processing, computer vision, neural networks and cognitive science. It has been widely used in authentication, human-computer interaction, vision communication and public security archives management, which shows an extensive application value. More and more researchers are devoting themselves into the area, producing a large numbers of new algorithms of practical application value, which is pushing the rapid progress of face detection and recognition technology.This paper put forward four new Haar-Like T features and applied the new features in the field of face detection and recognition. The details of the face detection and recognition algorithms based on Haar-Like T features were introduced in this paper and a serious of testing experiments were performed to illustrates the validity and superiority of the algorithms. For the key technologies involved in this article we described in detail and we also applied the algorithm proposed in this paper in the specific application development, which further illustrate the superiority of the algorithm in this paper. The main research work of this paper includes several aspects as follows:1. Four kinds of new Haar-Like T features were put forward in this paper and a serious of testing experiments on the new face detection algorithm based on Haar-Like T features were performed. Under the same experimental conditions a number of face positive samples and non-face negative samples were put into the Adaboost classifier, obtained two cascade classifiers. The detection experiment of the new algorithm was performed and141images containing599face were downloaded randomly from the Internet. The experiment compared the detection performance of T classifier with Haar-Like classifier and LBP classifier released by the latest OpenCV2.4.8. The experimental results showed the effectiveness of the new feature based face detection algorithms. The recognition accuracy and the recognition speed of the T classifier was better than Haar-Like classifier and LBP classifier. Compared with the latest face detection algorithms, the detection rate performed better, which showed the superiority of the new proposed algorithm. 2. Researches were did on the new Haar-Like T features, the new proposed features were applied to the field of face recognition. The T classifier trained above was used to extract features on the AR face database. The extracted features were put into the sparse representation classification (SRC) face recognition algorithm for face classification and recognition. The experimental results showed the effectiveness of Haar-Like T feature based face recognition algorithms. Testing experiments were performed on the AR face database and achieved a higher face recognition rate compared with the traditional face recognition algorithms such as NN, NS, SRC, SVM and so on, which further showed the superiority of the Haar-Like T feature based face recognition algorithm.3. This paper carried on the detailed instructions for some key technical problems involved in the process of face detection. Firstly introduced the instructions of some development tools such as Cmake、OpenCV、Visual studio2010. Then introduced the detailed process of how to compile OpenCV using Cmake. At the last of the forth chapter it showed the C++code of Haar-Like T features, the full process of training classifier and debug the key program.4. This paper made a secondary development to realize the function of face detection based on the iDS-2DF1-77A network high-definition infrared automatic track ball combined with OpenCV and MFC. The last chapter made a simple introduction for some key technology in the process of our developments. This paper achieved the application development based on the iDS-2DF1-77A network high-definition infrared automatic track ball and realized the face detection successfully.