基于车检器数据的高速公路事件检测可靠性提升技术研究

发布时间：2018-04-23 12:20

本文选题：高速公路 + 交通事件自动检测　；参考：《重庆大学》2015年硕士论文

【摘要】：事件检测是高速公路运营管理的重要支撑手段,其检测效果依赖于从传感器获得的大量实时数据,在实际应用中,由于数据质量的不可靠常常导致事件检测系统出现不可预见的误报警、重复报警以及漏报警等问题,直接影响系统工作的可信性。目前,学术界和工程界对高速公路事件检测的数据可靠性研究尚比较缺乏,不能满足实际应用需求。为此,深入研究各种数据不可靠因素对高速公路事件检测的影响规律,进而探索检测效果提升的方法,提高事件检测结果的可信性,将为高速公路运营事件管理提供更可靠的感知信息,促进高速公路运营管理水平提升,具有重要的实际意义。论文针对上述高速公路事件检测数据可靠性问题,以固定车辆检测器数据为典型案例,从分析数据不可靠因素对事件检测性能的影响着手,探索事件误报警、重复报警以及漏报警等产生的原因,进而研究高速公路事件检测可靠性提升的方法。主要工作如下:①进行了数据不可靠因素的影响分析。首先,通过分析事件检测系统的典型数据不可靠因素,发现对检测效果的影响主要源于数据缺失、数据异常和数据不同步三个方面。然后,针对数据缺失建立了事件检测视角下的数据完整性评价方法,针对数据异常提出了基于统计学及交通流理论的正确性评价方法,针对数据不同步问题,借助交通仿真软件Transmodeler及其GISDK开发工具,对不同步的机理进行了定性及定量刻画。最后,以经典加州算法为例,分析了三种数据不可靠因素对事件检测误报、重复报及漏报的不同影响程度,为后续工作奠定了基础。②分别从数据预处理和算法层面提出了事件检测系统可靠性提升方法。针对数据缺失和数据异常,提出了一系列数据筛选及修复机制,提升了检测算法对缺失数据和异常数据的容忍性;在算法上,利用相对漂移时间,提出一种面向数据不同步的改进加州算法,实际应用证明,在存在数据不同步的条件下,该方法与经典加州算法相比,检测率提高6.25%,误报率降低15.24%,漏报率降低6.25%。最后,将上述数据预处理方法和改进算法采用VS2010进行软件实现,并实际应用于重庆高速公路事件检测系统,使用结果表明:与经典加州算法相比,检测率提高6.66%,误报率降低21.54%,漏报率降低6.66%。因此,本文的研究成果能够有效降低数据不可靠因素对高速公路事件检测系统的负面影响,提高高速公路事件检测结果的可信性和可靠性。
[Abstract]:Event detection is an important supporting method for highway operation management. Its detection effect depends on a large number of real-time data obtained from sensors. The unreliability of data quality often leads to unexpected false alarm, repeated alarm and leakage alarm in the event detection system, which directly affects the credibility of the system. At present, the research on the reliability of highway event detection data in academic and engineering circles is still lacking, which can not meet the practical application needs. Therefore, the influence of various data unreliable factors on expressway event detection is deeply studied, and the methods of improving the detection effect are explored to improve the credibility of the event detection results. It will provide more reliable perceptual information for expressway operation event management and promote the improvement of expressway operation management level, which has important practical significance. Aiming at the reliability problem of the above expressway event detection data, this paper takes the fixed vehicle detector data as a typical case, from the analysis of the impact of unreliable data on the event detection performance, explore the event false alarm. The causes of repeated alarm and leakage alarm are discussed, and the method of improving the reliability of expressway event detection is studied. The main work is as follows: 1 has carried on the data unreliable factor influence analysis. Firstly, by analyzing the typical data unreliable factors in the event detection system, it is found that the influence on the detection effect is mainly from three aspects: missing data, data anomaly and data synchronization. Then, the evaluation method of data integrity from the view of event detection is established for the absence of data, the correctness evaluation method based on statistics and traffic flow theory is proposed for data anomaly, and the problem of data synchronization is discussed. With the aid of the traffic simulation software Transmodeler and its GISDK development tools, the asynchronous mechanism is characterized qualitatively and quantitatively. Finally, taking the classical California algorithm as an example, three kinds of data unreliable factors are analyzed, which have different influence on the false alarm, repeat report and false report of event detection. It lays a foundation for the following work. 2. The method of improving the reliability of event detection system is put forward from the aspect of data preprocessing and algorithm respectively. A series of data filtering and repairing mechanisms are proposed to improve the tolerance of the detection algorithm to missing and abnormal data, and the relative drift time is used in the algorithm. An improved California algorithm for data synchronization is proposed. The practical application shows that the detection rate is increased by 6.25%, the false alarm rate is reduced by 15.24%, and the false alarm rate is reduced by 6.25%, and the false alarm rate is reduced by 6.25% compared with the classical California algorithm. Finally, the above data preprocessing method and improved algorithm are implemented by VS2010 software, and applied to Chongqing Expressway event Detection system. The results show that: compared with the classical California algorithm, The detection rate was increased by 6.66, the false alarm rate was reduced by 21.54, and the false alarm rate was reduced by 6.66. Therefore, the research results of this paper can effectively reduce the negative impact of unreliable data on the expressway event detection system, and improve the reliability and credibility of the expressway event detection results.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;TP274

【相似文献】