面向交通拥堵的车辆鲁棒检测及分车道到达累计曲线估计
发布时间：2018-10-21 09:03
【摘要】：针对大交通量拥堵情况下现有视频车辆检测技术不能有效处理车辆相互遮挡而导致的大量漏检问题,提出了一种面向交通拥堵的车辆鲁棒检测及分车道到达累计曲线估计方法.首先,完成非拥堵区域的检测,避免针对交通拥堵停驶车辆进行复杂遮挡处理及检测的工作;然后,基于假设生成和验证框架,融合Ada Boost分类器与车底阴影检测结果,得到车辆鲁棒检测结果;最后,使用投影畸变车辆稳定特征将车辆位置划归特定的车道,准确估计分车道车辆到达累计曲线,实现针对交通检测断面分车道详细交通参数的有效分析.实验结果表明:该方法能够在高峰时段的交通拥堵状态下实时进行车辆鲁棒检测并准确地获取交通参数,有效避免针对车辆遮挡的复杂处理过程,对解决车辆到达率和车头时距调查成本高、工作量大、不确定因素多等问题具有实际的意义.
[Abstract]:In view of the fact that the existing video vehicle detection technology can not effectively deal with a large number of missed detection problems caused by mutual occlusion of vehicles under heavy traffic congestion, a method for robust detection of vehicles and estimation of cumulative curve of lane arrival for traffic jams is proposed in this paper. First of all, the detection of non-congestion areas is completed to avoid complex occlusion processing and detection for traffic jam and stop vehicles. Then, based on the hypothesis generation and verification framework, the Ada Boost classifier and shadow detection results under the vehicle are fused. The vehicle robust detection results are obtained. Finally, the vehicle position is assigned to a specific lane by using the projective distortion vehicle stability feature, and the cumulative curve of vehicle arrival is estimated accurately. To realize the effective analysis of detailed traffic parameters for traffic detection section divided into lanes. The experimental results show that the proposed method can detect the vehicle robust and obtain the traffic parameters accurately in the rush hour traffic congestion, and avoid the complex processing process of vehicle occlusion effectively. It is of practical significance to solve the problems of high cost, heavy workload and many uncertain factors in the investigation of vehicle arrival rate and headway distance.
【作者单位】： 北京工业大学城市交通学院;交通工程北京市重点实验室(北京工业大学);北京市城市交通运行保障工程技术研究中心(北京工业大学);廊坊师范学院计算机系;北京工业大学建筑工程学院;北京交通大学电气工程学院;
【基金】：国家自然科学基金资助项目(61573030,61511130044,61531005) 河北省高等学校科学技术研究青年基金资助项目(QN2015209)
【分类号】：TP391.41;U491
，
本文编号：2284635
[Abstract]:In view of the fact that the existing video vehicle detection technology can not effectively deal with a large number of missed detection problems caused by mutual occlusion of vehicles under heavy traffic congestion, a method for robust detection of vehicles and estimation of cumulative curve of lane arrival for traffic jams is proposed in this paper. First of all, the detection of non-congestion areas is completed to avoid complex occlusion processing and detection for traffic jam and stop vehicles. Then, based on the hypothesis generation and verification framework, the Ada Boost classifier and shadow detection results under the vehicle are fused. The vehicle robust detection results are obtained. Finally, the vehicle position is assigned to a specific lane by using the projective distortion vehicle stability feature, and the cumulative curve of vehicle arrival is estimated accurately. To realize the effective analysis of detailed traffic parameters for traffic detection section divided into lanes. The experimental results show that the proposed method can detect the vehicle robust and obtain the traffic parameters accurately in the rush hour traffic congestion, and avoid the complex processing process of vehicle occlusion effectively. It is of practical significance to solve the problems of high cost, heavy workload and many uncertain factors in the investigation of vehicle arrival rate and headway distance.
【作者单位】： 北京工业大学城市交通学院;交通工程北京市重点实验室(北京工业大学);北京市城市交通运行保障工程技术研究中心(北京工业大学);廊坊师范学院计算机系;北京工业大学建筑工程学院;北京交通大学电气工程学院;
【基金】：国家自然科学基金资助项目(61573030,61511130044,61531005) 河北省高等学校科学技术研究青年基金资助项目(QN2015209)
【分类号】：TP391.41;U491
，
本文编号：2284635
本文链接：https://www.wllwen.com/kejilunwen/daoluqiaoliang/2284635.html
