基于OpenHarmony的摄像机车牌识别技术研究

打开文本图片集

中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2026）05-0119-07

Research on License Plate Recognition Network of Cameras Based on OpenHarmony

CHANG Shaohui¹， LIN Yuan²， HUANG Yan'， SUN Yujing'，WANG Lingjiao'，LI Shaorong （1.Yunnan TransportationResearch Institute Co.，Ltd.，Kunming 65oo11，China; 2.Yunnan Communications Investment and Construction Group Co.，Ltd.，Kunming 65o1o3， China）

Abstract：Against the backdrop of the rapid development of intellgent transportation，to address the problems of highcros-platform deploymentcosts and low recognition accuracy incomplex environments for traditional license plate recognition，alicenseplaterecogitionsolutionwiththecordinationofsystem，hardwareandalgorithmisonstructedbased onOpenHarmony.This solution develops acamera system scheduling module，adopts theRetinex theory to optimize light compensation，uses CycleGANtogeneraterainandfogsamples，andcombineschannelpruningandquantizationtechnologies torealize the lightweighting of thealgorithm.Theresultsshow thatthe videostreamcrossdevice schedulingdelay is ⩽127ms ， the recognition rate under low illumination reaches 92.3% ，the rain and fog positioning success rate stands at 88.7% ，the highspeed vehicle capture rate hits 99.6% ，the algorithm volume is compressed by 73.3% ，and the inference time is ⩽37 ms， with allidicators metingthe design objectives.Thisresearch providesalow-costand high-robustnesstechnicalpathfortheedge deployment of intelligent transportation.

Keywords： License Plate Recognition Network; Retinex theory; CycleGAN; algorithm lightweighting

0 引言

近年来，云南省高速公路总通车里程已突破10000km ，基建建设进程放缓，已建路从“能通全通”进入“智慧化提升改造”阶段。（剩余9685字）