Research On Face Detection And Tracking Algorithms Of Humanoid Robot Vision

Research and applications on robotic field have been extended at present time, and the humanoid robots have more abroad application foreground because of they having the same appearance and characteristic with human, having human intelligence and being flexible as human. They can communicate with human, finish every kind of work such as mending, rescuing, detecting in hostile environment, and can play a vital role in amusement and service industry. So the research and application on humanoid robot is attended generally, and further more has been the most activity research hotspot in the intelligent robots field.Face detection and tracking is the key technology in face information processing field. It will have more expansive prospect in filed such as face automatic identification system, detection based on contents, vision monitoring and human-computer interaction and so on. As the first step of face identification, face detection has important role.As the research emphases, automatic face detection and tracking in complex background are performed. Firstly, four methods of face detection are researched, and RGB skin model is performed to used to abstract area which has the same color with human face. To excluding the influence to face detection, a linear means of illumination compensation is performed after analyzing some main means of illumination compensation. Proved by experiment that the effect of abstracting area has been improved largely, and that can meet the process demand.Secondly, the technology of image segmentation is researched. Block coloration method is used to segment several faces that maybe in the image and count the number of independent objects. Due to parting the image into block before process, the velocity of segmentation has been increase greatly. Then template matching is introduced through the idea of invariant moment. Ellipse fitting method for the face area is used to detect face, and distract the non-face area. At last, the number, the size and centroid of the face is obtained.Lastly, the direction of the face's movement is judged by comparing the faces'centroids in the two continuous images using difference method. Then single chip communicates with PC through serial port, and pass the word to the motor which drives CCD camera tracking the face real time. It has been found that the face detection and tracking system is reliability and real time in every kind of complex background and photoenvironment. The accuracy rate is above 90% and is sastisfactory.