Research And Implementation Of Face Detection SoC System Based On ARM Cortex-M3 Core

At present,the technology of face detection is becoming more and more mature,and it is widely used in aerospace,public security,civil life and other fields.Especially in the context of the outbreak,the contact-free face detection method is more in line with the requirements of safety,speed and convenience.With the development of the era of intelligence,more low power and high performance face detection systems are required.Based on the hardware platform of FPGA and the hardware-software co-design methodology,this paper implements the face detection SoC system on the basis of ARM Cortex-M3.The main works are as follows:1.By analyzing the workflow of the face detection system and adopting the software-hardware co-design method,the system architecture is defined and each functional module is divided.According to the functional characteristics of the system,the Xilinx Artix-7 FPGA development board with rich logic resources is used as the realization platform,and the OV7725 camera,DDR3 memory,LCD screen and other equipment are selected as the system peripherals.The design of image acquisition function,image cache function,and image real-time display function are completed respectively,which constitute the basic video image SoC system.2.Based on the video image function of the system,the video data is preprocessed by hardware acceleration.After the image is preprocessed by hardware acceleration such as graying,median filter,and binarization,the amount of data is reduced,and the speed of subsequent face detection and tracking is accelerated.According to the skin color of the face,the biological characteristics of the human eye and the proportion of the face,the hardware algorithm is used to screen and retrieve the face area.After the face is detected,the center coordinates of the face are calculated and the face area is marked and positioned.3.The system software part is completed and a face recognition SoC on the FPGA development board was built by combining hardware stem part.And the same time,board-level debugging and system function test are carried out.The results are shown that static or dynamic face targets can be quickly tracked in real time,which meet the expected requirements of the system.