Design Of Embedded Face Recognition System Based On ARM

Nowadays,the development of science and technology is changing with each passing day,and biotechnology is one of the most eye-catching ones.Traditional recognition technology can't meet people's needs.On this basis,fingerprint and voice recognition have been developed,and faces account for the largest proportion of visual information.Therefore,Developing face recognition technology is the future trend.With the rapid development of science and technology,some embedded processors are running faster than PCs,and embedded devices are characterized by miniaturization and high integration.Therefore,face recognition technology based on embedded platforms has become the current Research trends.This paper first introduces the development process of embedded system and image processing,and then introduces the common methods of face recognition in detail.Then it describes in detail the hardware platform of the embedded platform RK-3399,as well as the bootloader and Ubuntu system.The underlying environment,such as the embedded file system,is tailored,compiled,and ported.The face recognition algorithm mainly includes face detection algorithm and face recognition algorithm.Since the face detection and recognition algorithm training has a large amount of image processing,the TensorFlow framework is built in the Windows experimental environment under PC,respectively in WIDER FACE and CASIA-WebFace face database completes model training for face recognition and face detection algorithms.Firstly,the face recognition function is completed on the PC.The face target detection is detected by MTCNN and the face candidate area in the video is extracted.Face net extracts the feature vector and compares it in the face of the library.Finally,it is deployed on the RK-3399 embedded platform,and images are captured by a USB camera,and the image is grayed out and noise removed.The MTCNN algorithm completes the Face Detection Face net algorithm to complete face recognition,which is further improved compared with the traditional PC-based face detection recognition,which makes the face recognition system more convenient and more effective.After the design of the embedded face recognition system is completed,this paper tests the performance indicators such as the recognition rate and running speed of the system,and implements a simple person on the RK-3399 embedded platform using the traditional face recognition library that comes with Opencv.The face recognition system tests the two methods in the case where the illumination conditions and postures are various and the face is partially occluded.The results show that the ARM-based embedded face recognition system can operate stably and with high recognition rate,and has a good application prospect.