【作者】 杨博；

【导师】 金先龙；

【作者基本信息】 上海交通大学 ， 机械设计及理论， 2010， 硕士

【摘要】 随着我国城市化进程加快,机动车数量迅速增加,交通事故发生频率高居不下。通常,交通事故多发地一般位于高速公路、交叉路口等交通流量较大的地区。如果现场勘测时间过长,就可能造成道路阻塞、交通秩序混乱等后果。因此,要求现场工作人员能高效、准确地进行信息采集,并依据这些信息对事故过程做出科学推断。传统的事故现场勘测主要依靠手工定位和测量,工作效率低、测量精度不高、无法对事故场景进行二次验证。全站仪、激光测绘仪等先进设备在事故勘测中的应用在一定程度上弥补了传统方法的不足,但是这些设备一般较为笨重,且价格昂贵,普及率不高。与上述方法相比,本文所采用的摄影测量方法较为便捷、经济,测量精度较高,甚至可以不亲临现场,直接通过照片进行现场信息采集。摄影测量结果还可与各种交通事故再现方法相结合,起到提供初始信息和验证数据的作用。对交通事故现场进行摄影测量时,通常需要在现场布置标志点,并在图像上提取标志点作为摄影测量中的控制点或匹配点。目前,在对交通事故进行摄影测量的过程中,主要依靠手动或半自动方式来识别和定位事故现场摄像中的标志点。上述两种方式在一定程度上依赖于人工干预,未实现标志点识别与定位过程的全自动化。某些情况下,当摄影测量精度要求较高时,需要在三张以上图像上选取和匹配大量标志点时,依赖人工干预的选点方式将显得效率低下,而且人工选择标志点使得定位精度随机性增大。针对上述现状,本文对标志点的自动识别算法进行了研究与实验,节约大量工作时间并降低由于手工选点造成的误差,提高了交通事故现场勘测的自动化程度。另外,本文对事故现场摄影测量方法进行了研究。目前,国外已将摄影测量技术应用于交通事故现场图的绘制中,但主要采用航拍形式,不符合我国国情。国内偏重于数字化近景摄影测量在交通事故中的应用研究,并取得了一定成果。其中,以直接线性变换法为理论基础,使用普通的非量测数码相机对事故现场进行摄影测量和三维重建较为常见。该方法优点在于现场拍摄时对相机型号、配置等不做要求;不足在于必须在拍摄现场布置控制场,后期处理繁琐,测量精度低。本文在对传统摄影测量方法进行深入研究的基础上,针对直接线形变换法的不足,对现有方法予以改进,提出一种新型的交通事故摄影测量方法,并对其中涉及到的现场摄影方法和控制场布置测定进行了方法探讨和总结。最后,本文在标志物识别和交通事故摄影测量方法研究基础上,首先对2起真实交通事故案例进行了摄影测量分析,为事故现场图的绘制以及事故再现提供了所需二维和三维几何信息;接着提出现场图实时测绘系统的概念,初步建立了系统实现方案和系统软件系统的基本功能,并通过3组实验对软件样机主要功能进行了测试。更多还原

【Abstract】 Along with our country social economy development, vehicles increased rapidly, which led to an increase in traffic accidents. Usually, a traffic accident happens in an area with large traffic volume, such as freeway, road infall and etc. If the time to handle the accident is too long, it maybe induce road block and traffic out of order. As a result, the accident scene survey should be done accurately and efficiently. Traditional accident scene survey still relies mainly on the artificial measurement, which is low-efficient and low-precision and unable to carry on for two times to the accident scene.New technologies, such as Total Station, laser metering equipment and so on have make up these deficiencies. However, digital photogrammetry is more convenient and more economic compared with these techonologies. Scene information can be acquired from scene photos directly rather than from the real scene. And photogrammetry results can be applied as input and validated conditions in traffic reconstruction.In the traffic accident photogrammetry, calibration objects should be set in the scene to be as control points or reference points. At present, calibration points are recognized and located by handiwork or semi-automatic computer recongniton. The two methods relies on manual work in some extent. In certain conditions, such as selecting and marking large numbers of points on more than three photos, points selection relying on manual work is low efficiency. Aim to solve the problem, this paper studies on recognizing the calibration points of the photos automatically, by which measure time and errors have reduced a lot and traffic accident scene survey automation has improved.In addition, this paper has studied on the application method of digital photommetry in the accident scence survey. At present, Foreigner has applied photogrammetry techonology to draw the traffic accident scene map, using photos shot on the helicopter, which is unsuitable to our country situation. Our country s study focusd on the photogrammetry application methods in the traffic accidents scene survey, which DLT method using common digital camera has been studied by the most researcher. This method has no limits to the camera types, but demands setting controls in the scene and requires more time to process the related data. This paper proposes a new method used in traffic accident scene survey by improving DLT method , discusses and summarizes the related shooting methods and scene controls setting methods.Finally, based on the methods proposed before, the paper analyzed two real traffic accidents firstly and provided the necessary 2D or 3D geometric information for the traffic accident scene mapping and process simulation. And then, the paper proposed the concept of traffic accidents scene real-time mapping system , realized the basic functions of the system s software component and illustrated the software prototype by three tests.更多还原