摘要
视频监控系统视频处理实时性不高,有线视频监控系统迁移扩展性差、布线繁琐,低光照下视频图像获取质量和处理效果较差,针对以上问题,设计了一种低光照环境下基于FPGA+ARM的无线视频监控系统。首先在FPGA端初始化OV5640摄像头,使其输出RGB888格式的图像数据,图像数据经过格式转换缓存到DDR3中。然后提出基于边缘增强模块的改进RetinexNet图像增强算法,在HSV格式中利用高层次综合(High-level Synthesis,HLS)工具对亮度分量V进行硬件加速,在ARM端对亮度分量S进行线性拉伸,并完成低光照图像增强任务。在此基础上,通过5G WiFi模块发送图像数据,上位机显示器通过5G网络接收图像数据并实现HDMI接口输出。实验结果表明,系统针对昏暗图像改善图像细节,具有良好的显示效果,而且处理一帧图像耗时为136.08 ms,相比ARM和PC机,具有处理速度快、便携性高等优点。
Aiming at the problems existing in video monitoring system,such as low real-time video processing,poor migration scalability,complicated wiring,and poor quality of video image acquisition and processing effect in low light,a wireless video monitoring system is designed based on FPGA+ARM in low light environment.Firstly,the OV5640 camera is initialized in FPGAto output the image data in RGB888 format.The image data is cached in DDR3 through format conversion,and then an improved RetinexNet image enhancement algorithm based on edge enhancement module is proposed.In HSV format,the brightness component V is hardware accelerated by high level synthesis(HLS) tool,the brightness component S is linearly stretched in ARM,and the low-light image enhancement task is completed.On this basis,the image data is sent through the 5G WiFi module,and the upper computer display receives the image data through the 5G network and realizes HDMI interface output.The experimental results show that this system can improve image details for dim images,and has good display performance.Moreover,it takes 136.08 ms to process one frame of image,which is faster and more portable than ARM and PC.
作者
邓健峰
申利民
丁荔
李婧瑶
Deng Jianfeng;Shen Limin;Ding Li;Li Jingyao(School of Electrical Engineering,Guangzhou Railway Polytechnic,Guangzhou 511300,China;School of Mechanical Engineering,Guangzhou Institute of Technology,Guangzhou 510075,China)
出处
《机电工程技术》
2024年第7期137-140,176,共5页
Mechanical & Electrical Engineering Technology
基金
广州铁路职业技术学院新引进人才科研启动项目(GTXYR2314)
广东省普通高校类科研项目(2021KTSCX269)
广州铁路职业技术学院创新创业项目(2021SCSYS002)。
作者简介
邓健峰(1992-),男,博士,讲师,研究领域为深度学习、嵌入式技术;申利民(1978-),男,硕士,副教授,研究领域为深度学习、嵌入式技术;丁荔(1995-),女,博士,讲师,研究领域为深度学习、嵌入式技术;李婧瑶(1988-),女,博士,讲师,研究领域为深度学习、机器人故障诊断。
