【作者】 褚洪君；

【导师】 屈桢深；

【作者基本信息】 哈尔滨工业大学 ， 控制科学与工程， 2009， 硕士

【摘要】 随着我国城市化进程的不断加快,城市公路交通系统的压力不断加大,交通违章停车这一交通问题日益突显出来,对交通正常秩序的维护造成了非常恶劣的影响。本文基于视觉技术对违章停车的检测方法进行了研究,同时设计和实现了车辆违章停车自动抓拍系统。在对系统进行了现场测试的基础上对系统的性能进行了详细的分析。论文的主要内容如下:首先,论文对监控场景中运动车辆动态检测问题进行了研究。在分析比较目前比较常用的几种检测方法的基础上,应用了一种基于时间的运动历史图像的方法。这种方法可以充分利用监控视频携带的信息,对运动目标的表达独立于处理系统的性能,有较强的实时性。同时,论文在这一部分利用了一种基于阴影特性的阴影消除方法对与车辆一同被检出的车辆阴影进行了消除,实现了对运动车辆的准确检测。其次,论文对运动车辆的跟踪方法进行了研究。论文详细论述了跟踪匹配原理,并比较分析了几种常用跟踪方法。针对论题的应用,本文应用了特征点光流估计匹配的跟踪方法。这种方法实时性强,对车辆粘连、遮挡问题有一定的鲁棒性,并且应用这种方法很容易对车辆的违章行为进行判断,实现了在监控场景下的有效跟踪。再次,论文设计了高速球型摄像机精确定位算法。本文结合了线性摄像机成像模型及高速球型摄像机通信协议,对跟踪的违章车辆目标实现了二次定位,最终利用车牌识别软件可以完成对车牌的识别。并且在定位过程中可以获取了车辆违章的证据,从而完成了对违章车辆的处理。最后,在前述算法基础上本文设计了车辆违章停车自动抓拍系统。本文对系统的总体构架、各模块的设计、性能指标的评估方法进行了论述。并且论文还在不同的天气状况下,在实际场景中对该系统进行了实验,并从系统准确性、实时性、稳定性等方面对系统进行了评估。实验结果表明,该系统是车辆违章停车问题一种切实有效的解决方案。更多还原

【Abstract】 With the development of city and economy, the burden of the road transportation system becomes heavier and heavier. Illegal parking has become a more serious problem which has a very bad affection on normal transportation order. The paper is devoted to detecting method of illegal parking based on vision technology and illegal parking automatically snapshot system is designed and realized. What is more the paper analyzes the performance of the system in detail on the basis of field testing. The main contents of the paper are listed as follows:Firstly, this paper focuses on moving vehicle detection problem in dynamic scenes. A method named time-based motion history image is applied after analyzing and comparing several normal used methods. This method sufficiently uses the information surveillance video carries and the moving object is depicted independent of processing system. As a result this method has good real-time performance. Meanwhile this paper applies method based on shadow character to remove vehicle shadow which makes the moving detection more precise.Secondly, this paper researches on moving vehicle tracking method. Matching theory and several normal used methods are introduced. According to this specific application in this paper characteristic points optical-flow estimation matching tracking method is used. It has good real-time performance and is robust for vehicle adhesion and shading. Furthermore this method makes it convenient to judge vehicle behavior which realizes effective tracking in surveillance scene.Thirdly, this paper designs accurate location algorithm based on high speed dome camera. This algorithm combines linear camera imaging model and communication protocol and locates peccancy vehicle twice. Then the vehicle plate can be recognized by using license plate recognition software. In this process peccancy evidence image can also be achieved, then the processing to peccancy vehicle is over.Finally, on the basis of all these algorithms above, automatic snapshot system for illegal parking is designed. The frame of the system design of each module and estimation method of performance index are described specifically. Under different kinds of weather circumstance the system is tested in real scene. Then, the system performance including accurate capability real-time capability and stability is estimated. The result shows that this system is an effective solution for illegal parking problem.更多还原