【作者】 王伟；

【导师】 王小鹏； 刘立群；

【作者基本信息】 兰州交通大学 ， 电子与通信工程（专业学位）， 2020， 硕士

【摘要】 随着铁路事业日新月异的发展,一大批客运专线陆续投入使用,方便了人们的出行,但同时列车的运营安全也越来越受到人们的重视。在众多影响列车安全运营的因素中,异物入侵因突发性强、不可预见性高的特点显得尤为突出。传统的异物入侵检测方法因铁路线路跨越地域广、沿线自然环境变化等因素的影响会出现误检、漏检问题,且难以在禁行区域预先检测出异物使得列车司机没有足够多的反应时间,因此可靠性难以保障。针对上述不足,本文提出了一种铁路周界异物入侵检测方法。首先对周界区域给定约束条件;其次根据基于运动目标自适应检测的改进ViBe算法精确分割出运动目标前景;然后对分割的前景二值图像面积做加权处理并判断是否为入侵周界的异物;最后为了获得异物在周界内的实时运动状态可通过基于特征匹配的Kalman滤波跟踪算法对周界区域内异物目标实时跟踪。本文以提高铁路周界异物入侵检测的准确率为目标,主要研究内容体现以下几个方面:(1)在前景分割阶段,针对传统ViBe算法检测结果存在Ghost区域,且受环境变化影响在提取前景时容易产生误检或漏检等问题,提出了一种基于运动目标自适应检测的改进ViBe算法。首先在背景模型初始化过程中,通过对均值背景建模设置调节参数方式获取真实背景,利用该背景初始化ViBe背景模型;其次在前景检测过程中,根据场景变化引入自适应半径阈值对前景进行自适应检测;最后对检测结果中存在的空洞进行数学形态学闭运算填充。仿真结果表明,本文方法能够有效抑制Ghost区域,并在高动态环境下能够较完整检测出目标前景。(2)在目标跟踪阶段,为了避免目标跟踪时出现的误跟踪和漏跟踪问题,提出了一种基于特征匹配的Kalman滤波跟踪算法。首先对检测出的目标建立Kalman滤波跟踪模型;其次以Kalman滤波算法对目标在下一帧中的位置进行预测;最后通过匹配策略来确定匹配目标。仿真结果表明,本文方法对单目标和多目标的实时连续跟踪均具有较强的性能。(3)在目标前景分割算法和目标跟踪算法研究基础上,设计了一套铁路周界异物入侵检测方法。首先详细阐述周界区域的标识过程;其次设定异物的判决条件;最后构建铁路周界异物入侵检测总体流程。以拍摄的铁道视频做仿真实验,结果表明,相比于传统异物入侵检测方法本文方法检测准确率达到95%以上。更多还原

【Abstract】 With the rapid development of the railway industry,a large number of passenger dedicated lines have been put into use,which has facilitated people’s travel,but at the same time,the safety of train operation has also been paid more and more attention.Among the many factors that affect the safe operation of trains,foreign object intrusion is particularly prominent due to its sudden and high unpredictability.Traditional methods of foreign object intrusion detection can cause false and missed detection due to the influence of factors such as the wide span of railway lines and changes in the natural environment along the lines.Moreover,the failure to detect foreign object in advance in the forbidden areas makes the train drivers have insufficient response time.Therefore,the method reliability is difficult to guarantee.In view of the above deficiencies,this dissertation proposed a detection method of foreign object intrusion in railway perimeter.Firstly,the perimeter area is given constraints.Secondly,the moving object foreground is accurately segmented according to the improved ViBe algorithm based on adaptive detection of moving object,and then the segmented foreground binary image area is weighted to determine whether it is a foreign object intruding the perimeter.Finally,to obtain the real-time movement state of foreign object in the perimeter,the intruding foreign object in the perimeter area is realized by the Kalman filter tracking algorithm based on feature matching.This dissertation aims to improve the accuracy of foreign object intrusion detection in the perimeter of the railway.The main work is as follows:(1)In the foreground segmentation stage,aiming at the existence of Ghost area in the detection results of traditional ViBe algorithm,and being affected by environmental changes,it is easy to cause false detection or missed detection when extracting the foreground,an improved ViBe algorithm based on adaptive detection of moving object is proposed.Firstly,during the initialization of the background model,obtain the real background by setting the adjustment parameters for the mean background modeling,and use the background to initialize the ViBe background model.Secondly,in the process of foreground detection,an adaptive radius threshold is introduced to conduct the foreground adaptive detection according to the scene changes.Finally,mathematical morphology close operation is used tofill the holes in the detection results.The simulation results show that the proposed method can effectively suppress Ghost area and detect the foreground object in a high dynamic environment.(2)In the object tracking stage,in order to avoid the problems of false tracking and missing tracking in object tracking,a Kalman filter tracking algorithm based on feature matching is proposed.Firstly,a Kalman filter tracking model is established for the detected object.Then,the Kalman filter algorithm is used to predict the position of the object in the next frame.Finally,the matching object is determined by the matching strategy.The simulation results show that the proposed method is robust to real-time continuous tracking of single and multiple objects.(3)After the object foreground segmentation algorithm and object tracking algorithm are proposed,a set of intrusion detection method for foreign object in railway perimeter are designed.Firstly,the identification process of the perimeter area is explained in detail.Secondly,the judgment conditions for the foreign object are set.Finally,the general process of foreign object intrusion detection in railway perimeter is constructed.Using the railway video taken as a simulation experiment,the results show that compared with the traditional foreign object intrusion detection method,the detection accuracy of the proposed method is more than 95%.更多还原