Design And Realization Of Embedded Face Detection And Tracking System

At present, Face detection and tracking is becoming a more and more active subject., it has a quite wide application and it can be used in lots of important places, such as intelligent monitoring, public security (criminal identification) security video conference, validation system, checking attendance system, man-machine communication system, medical, digital library, etc.Face detection and tracking system includes face detection positioning and face tracking. Face detection positioning finds the faces and the position of faces from input image and separates the faces from the background. Face tracking tracks face in a period of continuous video frame, on the basis of detecting faces in the before frame from video image, the subsequent frame continues catching the position and size of faces. Generally speaking, face tracking is the development of face detection, its essence is real-time dynamic face detection.The existing face tracking systems often realizes on PC. They have some shortcomings such as high cost and serious power . With the development of the embedded technology, embedded processor DSP has small size, high reliability and strong function, therefore, it becomes ideal choice for image processing systems. As a result, this paper designed and implemented the face detection and tracking algorithm on the C6000 series in DSP TMS320DM6446 platform.In this paper, we had a further research on AdaBoost face detection algorithm and CamShift face tracking algorithm. Then combined them and realized on pc firstly. After that we transplanted the algorithm into embedded platform, the DSP terminal of DM6446, and then packaged algorithm using xDAIS-DM, staying calling from ARM terminal; On the ARM terminal of DM6446, we developed system application to achieve video collection ,video display and so on, and we realized face detection and tracking application by calling the face detection and tracking algorithm module running on DSP side with VISA API . In this paper, by building a hardware development platform and configuring a software development environment, using codec framework, we realized a face detection and tracking system including some functions such as video collection, video display and face detection and tracking on DM6446. At last, we tested the system in detail. When the camera is fixed, the system can realize detection and tracking of single face. In consequent 200 frames, it ran steadily and its speed is to 25 fps.