摘要
为准确地评估车辆检测各种算法在实际环境复杂场景的运行优劣,提出一种基于半人工标定的视频检测交通参数基准数据获取算法。在充分采集各种交通应用场景不同环境的多段交通视频的基础上,该算法主要包括2个方面:1)基于半人工标定的车辆位置-时间真实数据(车辆行进轨迹数据)获取;2)基于车辆位置-时间真实数据研究计算多种交通参数准确数据的算法。该算法充分考虑现有视频检测的条件,以简易的方式获得真实的交通参数。经过与TRAFICON交通视频检测系统的比较可知,基于半人工标定方法获取的交通参数基准数据具有高准确性,这保证了该计算平台可以有效用于视频检测性能分析与评估。
In order to accurately assess video detection algorithms in the actual operation of complex scenes with bad Weather conditions, this paper proposes a truth data computation framework of traffic parameter based on semi-manual calibration for video detection. On the basis of the full collection of various traffic scenarios with different weather conditions, our algorithm includes two aspects: 1) Real vehicle location - time data (vehicle trajectory) acquisition based on semi-manual calibration; 2) The accurate data of a variety of traffic parameters are calculated based on the real data of the vehicle location. The algorithm takes into account the existing detection conditions in a simple way to obtain real traffic parameters. By comparison with the TRAFICON traffic video detection system, the proposed method in this paper can obtain baseline traffic parameters data with high accuracy, which ensures that it can be effectively used for performance analysis and evaluation of video detection.
出处
《中国科技论文》
CAS
北大核心
2015年第7期788-793,共6页
China Sciencepaper
基金
高等学校博士学科点专项科研基金资助项目(20111103120015)
关键词
交通工程
交通信息采集
机器视觉
基准交通参数
半人工标定
性能评估
traffic engineering
traffic information collection
computer vision
ground truth
calibration of traffic parameters
performance evaluation