Embedded Face Recognition System Based On ARM

With the rapid spread of the network and the continuous improvement of the transmission of information, people are enjoying the benefits of this huge change, but at the same time people also have been facing the information security threat as result of virtual interaction between people. So, to identify person’s identity in all kinds of information exchange platform, and to record people’s event information in public places, has been the primary means to deal with this threat. The face recognition technology has a wide range of market applications in video surveillance, target tracking and other fields, because it can collect sufficiently complex visual features in friendly and secret way. The face recognition technology requires many techniques to process images, such as image processing, physiology, anatomy and so on. In recent years, the recognition rate of it gradually increases because of the appearance of new algorithms and the improvement of old ones, with the rapid advancements of related technologies. Traditionally, we always implemented this technology on PC, so it can not be applied in diverse areas and meet front-end requirements of intelligent video captures because of its shortcomings such as low mobility, poor portability, high cost and so on. Meanwhile, due to the rapid development of embedded systems technology (ARM, DSP, etc), we have been able to make implementations of this technology with high performance and low cost. Therefore, research of the face recognition technology based on embedded systems techniques really has huge market demand.This article is about the design and the exploitation of the embedded face recognition system, studying from the execution of the embedded system in ARM in both its hardware and software. In terms of hardware, I completed the overall design of the system’s hardware platforms, including embedded microprocessor, LCD touch screen, USB camera, memory, etc.; in terms of software, I completed the embedded Linux operating system migration, device drivers porting and application software implementation. At first, we built embedded software development environment, including installation and configuration of cross-compiler and virtual machines; then I finished the migration of embedded Linux operating system, including Bootloader (u-boot) transplant, Linux kernel customization, the root file system installation and the associated driver transplant. In the application phase, first of all, I design and exploit graphical user interface module by Qtcreator; then achieve image acquisition module of system, following the programming architecture of Video4Linux2; At last, I implement the face detection module, feature extraction module and face matching module which were written in C language; Finally, I implemented the embedded face recognition system on ARM, by migrating above-mentioned modules to ARM.Finally, we designed a attendance system as the application of the face recognition system based on embedded systems techniques, the results of actual test show that the recognition reached 95.3% in the good light and the recognition reached 90.6% in the bad light, recognition time consuming can be limited to 1～1.5s.