Research On Anti-Halo Technology Of Heterogeneous Image Fusion Based On Multi-Core Heterogeneity

In the process of vehicle driving at night,the bright lights of coming vehicles affect the driver’s vision easily.The driver’s vision is blinding and he cannot accurately judge the road conditions ahead,resulting in potential road safety risks.The infrared camera and visible camera shoot halo scenes synchronously to carry out alien image fusion,which can effectively eliminate halo information in the fused image and reduce the adverse effects of halo on human eyes.However,the anti-halo image fusion algorithm is complex and computant-intensive,and it has some shortcomings such as low computational efficiency,long time and difficult application.Therefore,in this paper,the anti-halo image fusion algorithm based on ZYNQ heterogeneous platform is implemented for hardware acceleration,and an embedded anti-halo image fusion system with fast computing speed,high image processing efficiency,small size and low power consumption is designed to reduce the interference of halo to drivers and improve the safety of nighttime driving.The work of this paper is as follows:1)In order to ensure the synchronization of alien image data during image fusion,the design of parallel and synchronous acquisition of alien image is carried out.Firstly,a VDMA-based image synchronization acquisition project is designed.Then the Linux system is transplanted on the ARM end,and the driver design of USB3.0 industrial camera and VDMA image handling module is completed.The output module is designed on the FPGA end.Finally,synchronous acquisition and display of infrared and visible video images are completed under the heterogeneous system of ARM+FPGA.2)In order to improve the quality of anti-halo image fusion,the digital logic optimization design of image preprocessing algorithm is carried out in FPGA terminal of ZYNQ.Firstly,affine transformation with small amount of computation is used to achieve the registration of heterogeneous images,which eliminates the spatial differences of heterogeneous images.Then,multi-scale Retinex(MSR)image enhancement algorithm is used to enhance the dark detail information of infrared image and visible image.Since MSR involves convolution operation of convolution kernels of different scales,the complexity is relatively high when implemented by FPGA,so the calculation process is optimized.The minimal-scale convolution kernel is reused many times to complete the convolution operation of different scales,which greatly reduces the difficulty of implementing MSR in FPGA,and significantly enhances the detail information of image dark.3)In order to improve the brightness and clarity of the fused images of the embedded night vision anti-halo system,an NSCT image fusion algorithm based on YUV color space is designed on the FPGA terminal of ZYNQ.Meanwhile,in order to completely eliminate halo information from the fused images and avoid halo information from participating in the fusion process,the fusion rules of high frequency moduli and low frequency weight adaptive are designed.By utilizing the hardware advantages of FPGA rich logic resources and parallel architecture,the brightness and clarity of the fused image are significantly improved and the driving safety is effectively improved on the premise of ensuring the real-time performance of the anti-halo system.Vivado development platform was used to complete the integration of the anti-halo system hardware module,and the anti-halo system was verified and analyzed in multiple halo scenes.Multiple groups of experimental results showed that the multi-core isomerization based heterogeneous image fusion anti-halo system designed in this paper had high real-time performance,and the output anti-halo video image was smooth and clear.It has the advantages of high real-time performance,low power consumption and convenient application in the field of automobile safety.