面向停车场场景的多传感器融合匹配算法与融合数据的并行处理  

Multi-Sensors Calibration and Fusion Algorithm and the Fusion Data Parallel Processing Based on the Parking Area Scenario

作　　者：宝鹤鹏 陈超 王磊 BAO He-peng;CHEN Chao;WANG Lei(Automotive Data of China(Tianjin)Co.,Ltd.,Tianjin 300380)

机构地区：[1]中汽数据(天津)有限公司,天津300380

出　　处：《现代计算机》2020年第6期76-82,共7页Modern Computer

摘　　要：自动泊车是无人驾驶领域可率先实现量产的重要功能之一。实现自动泊车功能,首先面临着停车场数据的采集和分析工作。基于三种典型停车场场景,从硬件传感器搭建,传感器感知算法开发和场景数据处理分析三方面分别展开:对采集平台搭建工作,主要介绍如何进行传感器的安装、时间轴标定,和空间坐标统一;基于前视毫米波雷达和视觉传感器,重点介绍这两种传感器的匹配算法和目标级别的融合算法;基于场景提取与大数据处理的数据分析,开发自动化批量数据处理Python代码,实现TB级的原始数据标注,并在短时间内完成自动化批量处理,整个技术流程工具链具有较高的科研价值和商业价值。Automatic Parking is the most significant function that could achieving mass production in the field of unmanned driving.For the purpose of automatic parking function,we’re foundamentally facing the corresponding data collection and data analysis.In view of three classical parking scenarios,this paper is concentrating on three contents including the sensor hardware construction,the perception algorithm devel opment,and scenario data analysis.For the the hardware construction,we focus on the sensor installation,time calibration,and space unifica tion.Based on the front camera and RADAR data fusion,we put great emphasis on the multi-target matching and target-level fusion algo rithm.And for the purpose of the scenario extraction and big data analysis work,we are developing automatic big data processing Python code,realizing TB level original data labeled,and data completely processed in short time.The whole technology process and tool chain in this program,have extensively high level scientific research and commercial value.

关 键 词：多传感器标定方法 毫米波雷达与摄像头融合与匹配算法 大数据并行处理算法 停车场数据采集 低成本传感器硬件方案 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置] U463.6[自动化与计算机技术—控制科学与工程]