基于Zynq雾霾天气下的实时车牌识别系统

Real-time license plate recognition system based on Zynq haze weather

秋冬时节雾霾天气频发,严重影响了监控系统里的车牌识别精度,迫切需要能够进行实时处理的小型化的嵌入式车牌识别系统。图像去雾目前主要是在PC端进行,实时性低。本文提出一种基于Zynq雾霾天气下的实时车牌识别系统。通过改进去雾算法对颜色空间(HSV)中的亮度分量进行处理,然后通过中值滤波进行去噪,最后将处理后的图像利用车牌识别相关算法进行处理。该系统采用ARM+FPGA异构的方式,PS(ARM)端负责简单算法运算、IP核控制与驱动等,PL(FPGA)端负责需要大量并行运算的算法。通过实验结果对比,该系统能够达到预期的效果,具有实时性、可移植、可嵌入和功耗低等特点,能够满足在雾霾天气下的车牌识别的性能要求。

The frequent occurrence of haze weather in autumn and winter has seriously affected the license plate recognition accuracy in the monitoring system. There is an urgent need for a miniaturized embedded license plate recognition system that can perform real-time processing. Images dehazing is currently mainly performed on the PC side, with low real-time performance. In this paper, a real-time license plate recognition system based on Zynq haze weather is proposed. The luminance component in the color space(HSV) is processed by an improved dehazing algorithm, and then denoised by median filtering. Therefore, the processed image is processed by the license plate recognition algorithm. The system adopts the ARM+FPGA heterogeneous method. The PS(ARM) side is responsible for simple arithmetic operations, IP core control and driving, etc., and the PL(FPGA) side is responsible for the algorithms that require a large number of parallel operations. Through the comparison of experimental results, the system can achieve the expected effect, and has the characteristics of real-time, portable, embeddable and low power consumption. It is verified that the proposed system can meet the performance requirements of license plate recognition in haze weather.