Real-time object detection based on the heterogeneous SoC platform

In order to solve the problems of power consumption, portability, real-time and volume limitation of image acquisition and processing system based on PC, the object detection system based on SoC FPGA, in-built hard-core ARM processors, is implemented. By co-design of hardware and software based on SoC FPGA development platform and embedded Linux development, the system realizes the image acquisition of the CMOS sensor, storage of SDRAM, data communication of dual-port RAM and VGA display. In the meanwhile, ARM-based HPS controls dual-port RAM to read or write image data and the image pre-processing and background subtraction algorithm are implemented. Experimental results show that this system has the high accuracy of detection and the system achieves a frequency of 50 MHz reaching 19.8 fps with resolution 640 x 480 pixels and an estimated power consumption of 1.19 W. This system has the advantages of flexibility, high speed and good portability and it is a valuable reference for the realtime image acquisition and processing system. © 2017, ICPE Electra Publishing House. All Rights Reserved.