基于双目视觉的车道级高精地图制作

Generation of Lane-Level High-Precision Map Based on Binocular Vision

自动驾驶汽车工业对促进国际经济、社会、科技、安全发展,带动制造业转型升级及产业发展具有重要意义。高精地图作为实现汽车自动驾驶的关键基础设施,是交通资源全时空实时感知的载体和交通工具全过程运行管控的基础。项目利用配备双目视觉设备的移动测量车采集数据,通过基于神经网络的自动识别算法提取要素,使用DMI平台处理和检查数据,实现了低成本的车道级高精地图制作。文章重点说明了基于双目视觉的高精地图制作的技术流程以及各环节的技术要求,提供了在车联网示范项目应用实例,并指出面向大规模商业应用需要进一步研究的方向。

The auto driving industry is great significance for promoting the development of international economy,society,science and technology,security and promoting the transformation and upgrading of manufacturing and industrial development.As the key infrastructure to realize auto driving,high-precision map is the carrier of real-time perception of traffic resources in all time and space and the basis of the whole process operation and control of vehicles.In this project,the mobile measuring vehicle equipped with binocular vision equipment is used to collect data.The automatic recognition algorithm based on neural network extracts the elements,and the data are processed and checked by the DMI platform,so as to realize the production of lane level high-precision map at a lower cost.This paper focuses on the technical process of high-precision map making and the technical requirements of each link,provides an example of application in the Internet of vehicles demonstration project,and points out the direction of further research for large-scale commercial application.