摘要
提出一种基于量化神经网络的端到端车牌检测与识别系统,实现了车牌快速端到端的检测与识别。实验结果表明:所提出的网络可以有效地应用于车牌检测与识别,实现了最高99.2%的识别准确率与76 FPS的识别速度。相比浮点卷积神经网络参数量降低约32倍。
An end-to-end license plate detection and recognition system based on quantitative neural network is proposed,rapid end-to-end detection and recognition of license plates is realized.Experimental results show that the proposed network can be effectively applied to license plate detection and recognition,the maximum recognition accuracy of 99.2%and recognition speed of 76 FPS are achieved.Compared with floating-point convolutional neural network(CNN),the parameter amount is reduced by about 32 times.
作者
张旭欣
金婕
ZHANG Xuxin;JIN Jie(College of Electronic and Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201600,China)
出处
《传感器与微系统》
CSCD
2020年第12期103-105,共3页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61701295,61801286)。
关键词
端到端车牌检测与识别
卷积神经网络
量化神经网络
end-to-end license plate detection and recognition
convolutional neural network(CNN)
quantized neural network