基于yolov3的车牌定位识别系统

License plate location and recognition system based on yolov3

为解决目前市场上车牌识别设备的高时延以及对高性能处理器高依赖性的问题,提出了一种基于yolov3的车牌定位识别方法,实现对车辆的车牌进行自动定位与识别。该方法通过收集5000张车辆图像数据,打包为VOC数据集,使用LabelImg工具对图像进行标注,并构建13个卷积层的yolo模型,在loss曲线趋于稳定之后,完成模型的训练,最终使用已训练的模型对车辆图像数据进行车牌识别。实验表明:在1000个测试数据中,该方法识别率达97.5%,平均耗时18.19 ms,能够快速精准的对车辆进行车牌识别。

In order to solve the problems of high time delay and high dependence on high-performance processor of license plate recognition equipment in the market,a license plate location and recognition method based on yolov3 is proposed to realize automatic location and recognition of vehicle license plate.In this method,5000 vehicle image data are collected and packaged into VOC data set,the images are labeled with labelimg tool,and the yolo model of 13 convolution layers is constructed.After the loss curve tends to be stable,the model training is completed,and finally the trained model is used to recognize the vehicle image data.The experimental results show that in 200 test data,the recognition rate of this method reaches 97.5%,and the average time is 18.19ms.It can recognize the vehicle license plate quickly and accurately.