| Thanks to over thirty years of development in the study of face detection algorithm, many encouraging achievements have been acquired. Along with the development of the algorithms and expanding of face detection’s range of application, all kinds of system architectures of face detection targeting at different applications start with turning up. However, among these systems have the disadvantages, for example, the systems are inseparable, customized and expensive, and their processors need to be in possession of high performance, which undoubtedly block the development of face detection technology and its corresponding products.Modular design is taken as an effective notion of system design and has been successfully applied to the design of PC, resulting in that almost every household in the world has PC. New, the notion of modular design is helping the design field of robot implement a new reform. Face detection is an important part of robot-vision system. So, the modular design of face detection system not only is conducive to robot’s, but also can make face detection system eliminate the disadvantages mentioned above and obtain some new developments.Due to above reasons, from that point of modular design, this paper designs a face detection system in possession of the characteristic of modular system. This system is divided into two modules, the first module named as modular device of aiding face detection and the second module called as target system, and based on the two modules, what this paper does is as follows:(1) Select USB bus as standard interface of modular design, making the two modules mentioned above implement face detection by connecting the adopted USB bus interface. After that, adopt the integrated USB interface chip Cy7C68013A to implement USB communication protocol.(2) On the basis of the reasons why the proposed system consists of the two modules mentioned above, and the central processor of the first module is FPGA and the second module is general CPU, this paper proposes a face detection way which is hierarchical, coarse-to-fine and local-to-overall, and is composed of numerous face detection algorithms. First, the first level of face detection algorithm, skin tone-based and variation of gray level of eyes’image, which is easy to parallelly operate and in possession of simple formulas used to calculate is implemented on the FPGA, so as to fast determine whether there exists face within input image and the face’s position, second, based on the result of the first level of face detection algorithm, the second level of face detection algorithm, combining principal component analysis with support vector machine, which is complicated in the formulas used to calculate is implemented in general CPU, in order to detect face further.(3) Because the design procedure of USB interface is divided into two parts, USB function device and USB master device, the first module for using FPGA as central processor is seen as USB function device, resulting from that it takes advantage of merits of FPGA in the procedure of design of interface and parallel calculation. On the FPGA of the first module, the image capturing of CMOS digital image sensor, image frame’s storing of SDRAM-based, communication between USB interface chip and FPGA, and the first level of face detection algorithm mentioned in the (2) is implemented, resulting in that the first module becomes a device of aiding face detection in possession of the characteristic, Plug and Play, similar to USB. On the other hand, taking development cycle into consideration, the second module is implemented in general PC and take it as USB master device to finish the drive program’s configuration of USB interface chip, Cy7C68013A, and the development of corresponding application program. The main functions of application program contain reading image data, acknowledge signal of whether face exits and coordinate signals of face position, which are transmitted by the first module through USB interface, and then based on the acknowledge signal, determine whether application program continues to execute the second level of face detection algorithm mentioned in the (2) to detect further face.(4) By testing the whole face detection system, the acquired results show that the designed system can implement real-time face detection. Therefore, the system is a modular face detection system in possession of the features, commonality, flexibility and robustness. |
