Face Detection And Gender Identification Based On Video

With the development of computer technology, computer intelligence has spread over various areas of our lives. As a branch of computer intelligence, the technology of video detection, has a significant role in promoting the development of all areas of computer intelligence. With the popularity of intelligent video surveillance system, analyzing video information has become a priority. This paper focus on realizing face gender recognition for intelligent video surveillance system.The design of this system is from the cooperate project between Task Force and the State Grid Electric Power Research Institute in Shandong Province, This project needs to be do security work through substation inspection robots. Since the substation belongs to highly risk area, it has high safety requirements, thus to join the gender recognition functionality to enhanced inspection robot’s intelligent. This system can also be used in other cases, such as in large amusement, shopping malls, pedestrian malls, etc., computer automatic statistical gender composition of consumer groups, helping businesses provide more attractive services according to the gender composition of the population.Gender identity recognition technology is an important part of technology. In all aspects such as national security, crowd detection, it is playing an important role. Currently, there are many ways to gender discrimination, such as face, iris, fingerprints, footprints, and voice, etc. Compared with other biometrics, facial features are more natural, convenient and non-contact. Thus it is widely used in video surveillance, security and other fields. However, the process of video capture will be affected by the weather, lighting, video capture devices and other factors, which causes critical information of the video image are missing. Hence it increases the difficulty of gender identification from collected video images.The paper can detect the existence of faces, at the same time identify person’s gender. This paper proposes an algorithm which combines variance normalized and convolution to solve the uneven illumination in videos. The paper has developed a video-based facial image recognition system, which most effectively solve the current human face recognition problem. The main work of this paper includes the following sections:1) The first chapter summarizes the overall research on the history and development of the human gender identification and gives a brief description for all stages of the development process. It gives a summary on current way of Biometric identification, meanwhile analyzes as well as compares face-based image recognition and fingerprint recognition, palm recognition, iris recognition, etc. It emphasizes the advantages of face recognition technology based on gender and development prospects and significance.2) The second chapter describes the current mainstream face detection algorithm, focusing on the Adaboost algorithm and briefly described its development, detailing the implementation principle on Adaboost algorithm. It describes Haar features in detail, analyzing the feasibility of the use of Haar facial feature representation. Also, it gives details of how to use Haar features and a combination of Adaboost algorithm for face detection and extraction area.3) The third chapter investigates inequality lightening at home and abroad. It compares the main processing algorithms for mainstream, proposing a new lightening processing algorithms, and apply it to the face image. This algorithms combines variance normalized and convolution to solve the uneven illumination in post-gender discrimination.4) The fourth chapter gives a comparative analysis of the principal component analysis algorithm(Principal Component Analysis, PCA), linear discriminant analysis algorithm(Linear Discriminant Analysis, LDA), Active Shape Model(Active Shape Model, ASM) and active appearance model(Active Appearance Model, AAM). It also analyzes PCA and LDA on the principle of the algorithm is analyzed. Development and principles AAM model are described in detail, using the principle of AAM model, the extraction of a larger gender related characteristics were modeled to reduce gender recognition algorithm complexity. Finally, the SVM classifier principle of gender classification algorithms are described.5) Chapter V shows the video-based face recognition system’s main functions, and each function were introduced. In this system, the main use of Adaboost algorithm to extract the video face detection, and the detected face region is treated separately. By AAM modeling method, the gender-determining characteristics of calibration records, and then use SVM classification algorithm for classifying extracting features of gender, which determines character gender that the video image appears.