第8章交通应用建模科硕.pptVIP

实验与结果 唐炉亮 Tel 027Email：tll@ luliangtang@ * * 基于点的轨迹描述方法：每条轨迹由对应的一系列采样点组成，每个点包括时间，坐标，速度和角度信息 基于Segment的方法是将划分为一系列的线性子轨迹，记录每隔子轨迹的进入与驶出时间点，多个子轨迹段和一条路网路段相关联 * * we proposed an integrated Floating car Trajectory Model called Routes-Features FCD Trajectory Model. Every single trajectory is composed of a set of Features and Routes as well as a time tag and a car tag. The Features are the points abstracted from the original floating car data because of their special characters. The three types of features are O-D nodes, which means the origin and destination nodes of the trajectory; and velocity feature, which shows an apparent velocity change, this often appears in a traffic jam or signal waiting events. The third type of feature is the Activity features. They are the tracking points near some activity locations like school and stayed a period of time, this can be detected by a calculation using the certain high level activity discover algorithms A linear referencing system is used to describe the location information of all these three type of features. The Routes are all the Road Links the vehicle traveled along with two time stamps and along-road-speed description. By using linear referencing system and road link as route segment, the trajectory model is Traffic network embedded. * * 左上为原始数据，右上为按照速度特征压缩，左下为按照道路节点压缩，右下为速度特征+道路节点压缩 * * 第一个图是在交叉路口的匹配结果，第二个图是平行双线路（行车方向不同）的匹配结果，第三个图是稀疏点的匹配结果 * * * * * * * * * * * Making Road Maps from Trajectories Raw GPS traces Road map Crowdsource GPS traces from everyday vehicles 基于浮动车数据的车道信息获取 Brakatolas 将GPS定位测量值的规律描述为一个二变量标准正太分布的概率函数。 经验统计是取2倍σ内的值作为研究对象，但是对于浮动车数据而言却不能得到良好的结果。 采用Delaunay三角网进行粗差剔除以及点到中心线距离粗差去除得到经验概率：90% 车道宽度检测 同一条路径上的宽度探测示意图 目标路段 覆盖在目标路段上的原始浮动车数据 真实影像资料 车道宽度检测 基于贝叶斯原理的车道模式识别 贝叶斯分类器：贝叶斯分类器是一个统计分类器，其主要通过对对象的先验概率，利用贝叶斯公式计算出其后验概率，即该对象属于某一类的概率，选择具有最大后验概率的类作为该对象的所属的类。 目标路段x1 宽度w1 H为目标路段的为XX类型的假设，P（X|H）是建立在H基础上成立的概率，按照贝叶斯公式即可求出其值。 贝叶斯公式： 基于贝叶斯原理的车道模式识别 根据方向属性挑出路段上同向的文件_1.txt,逆向文件