摘要
为了保证全自动驾驶条件下列车编号识别的准确性,采用机器学习算法选择车辆编号区域图像的训练模型;根据车号位置、司机登车平台结构形式,合理选择车号机器视觉系统安装位置,确保所获取车号图像的完整性;采用YOLO算法对形态学处理完的图像进行字符识别,识别准确率大于99.74%,满足3σ原则,从而验证了训练模型以及算法的可靠性。
In order to ensure the accuracy of vehicle number recognition under fully automatic driving conditions,machine learning is used to train the vehicle number area selection.According to the height of the vehicle number,the position of the driver's boarding platform and the structure form,the mechanical vision installation position of the vehicle number is reasonably selected to ensure obtain the completeness of the car number image.Using the YOLO algorithm to perform character recognition on the morphologically processed image,the recognition accuracy is greater than 99.74 percent,which satisfies the 3 principles,thereby verifies the reliability of the training model and the algorithm.
作者
杨朋朋
Yang Pengpeng(China Railway First Survey and Design Institute Group Co.,Ltd.,Shaanxi Xi'an,710043,China)
出处
《机械设计与制造工程》
2023年第1期125-129,共5页
Machine Design and Manufacturing Engineering
基金
中铁第一勘察设计院集团有限公司2020年度科研(软件)开发项目(院科20-23)。