Research Of Face Detection And Recognition System For Mobile Robot

The aim of face detection and recognition is to judge whether faces exist, getting the amount and positions of faces and confirming the identities of the faces in different scenes. In the recent years, with the development of Pattern Recognition and Artificial Intelligence, face detection and recognition has become the important research subject promoted by the demand of airport security, video surveillance, et al. Meanwhile, the researchers in the mobile robots field has concerned on effective interaction between users and robots. For the mobile robots, acquiring information by vision is one of the most important sensing methods. Thus, developing face detection and recognition algorithms for mobile robots has good value and future prospects.The paper presents a novel real-time face detection and recognition system in the complex background. A multi-information inosculation method consisted of Adaboost algorithm and color information for face detection part is proposed. Adaboost algorithm is currently the most rapid face detection algorithm, which adapts to changes in illumination and posture to some extent. The skin color model in YCbCr space is also robust to different changes except illumination. It is employed to select the parts that may be not skin areas such as doors, windows or walls from the information detected by Adaboost algorithm. Embedded Hidden Markov Model (EHMM) is presented using 2D DCT feature vector as the observation vector to recognize the detected faces. A face should be described as a whole including not only each organ's numerical characters but also their various appearances and their relations. EHMM takes full use of face information to get the high recognition accuracy.The paper also establishes a mobile robot system with the function of face detection and recognition. The system consists of a controlling platform named of K-ICNP, a mobile robot and vision processing computer which are scattered distribution. The mobile robot transmits the video signal to the vision processing computer through wireless camera. The vision processing computer detects and recognizes faces from the video based on the algorithms above. Then the results will be transmitted to the platform through wireless network. The controlling platform will control the movements of the robot based on the detection results and talk to the people in front of the robot based on the recognition results.The effectiveness of the algorithm proposed in the paper is validated in the experiments using the mobile robot system. And the construction of the system has established the foundation for the further research.