Design Of Real-time Intelligent Recognition System For Vehicle Video Information Based On MPSoC

In the era of big data,massive amounts of video data are generated in surveillance systems in various application scenarios at all times.For the video vision system in the field of autonomous driving,the demand for smart camera terminals with small delay,high real-time performance,strong edge computing capabilities and independent hardware platforms arises at the historic moment.However,at this stage,research on vehicle-mounted smart camera terminals is focused on software algorithms,lack of independently developed high-performance hardware platform research and development.Besides,limited by hardware performance,processing delays are large,and the degree of marginalization is low.In view of the above-mentioned situation and project requirements,this paper designs a set of MPSo C-based vehicle-mounted intelligent camera system for road traffic automatic driving system.The system collects and processes multi-channel ultrahigh-definition video images,then detects,extracts and recognizes specific targets in the video,finally displays the ultra-high-definition video on the local display screen.The system uses ZYNQ Ultrascale+ MPSo C EV series chip as the core processor.The peripheral contains 4 ultra-high-definition image sensor interfaces and 1 HDMI 2.0ultra-high-definition video output.There are 6 pieces of 2400 Mbps DDR4 for off-chip storage on the board.The hardware platform of this system is a self-designed 14-layer printed circuit board,including 1159 electronic components,3871 connection networks and 18 power sources.There are 20 pairs of MIPI-DPHY signals with the highest transmission rate of 1.2Gbps and 96 DDR4 signals with a rate of 2400 Mbps.Therefore,in the hardware design stage,the entire system faces the challenges of signal integrity,power integrity,and electromagnetic compatibility.In this design,through theoretical calculation and software simulation,the above problems are solved smoothly,with the impedance strictly controlled,and the layout and wiring are reasonably operated.Based on the hardware platform descriptioned above,this article independently designed and implemented a set of real-time vehicle-mounted traffic sign recognition system,including FPGA programmable logic,ARM processing system,which are developed using Verilog HDL and C languages.FPGA program design is used to realize the functions of ultra-high-definition video data reception and decoding,video highspeed off-chip buffering,target detection and identification,local ultra-high-definition video display.ARM acts as a coprocessor to improve system programmability.It controls the working mode of the sensor,the AXI bus and the working mode of the HDMI transmitter software and hardware.In summary,the vehicle-mounted smart camera system designed in this paper realizes the collaborative processing of local software and hardware.The platform initiates image acquisition operations at the edge,and provides real-time target recognition and analysis application services at the data source.Non-control image and video data are processed and stored at the edge,shortening the response time of remote control,and eliminate hidden traffic safety hazards caused by insufficient response to road condition information.Therefore,the system has certain practical significance.