【作者】 王妍；

【导师】 任彦硕；

【作者基本信息】 东北大学 ， 控制工程（专业学位）， 2012， 硕士

【摘要】 车流量检测是智能交通监控系统的一个重要部分。通过对道路车流量的实时检测,得出交通状况和交通信息,并据此做出信号控制来指挥交通,从而可以减轻道路拥挤程度,提高交通设施的利用效率,保证整个交通系统的高效流通。基于视频的车流量检测以检测区域广、操作简单,提取信息全面等诸多优点越来越被重视,成为新兴的研究热点。通过对传统的车流量检测方法的技术特点和优劣势的分析与比较,本文采用一种针对局部图像的帧差法和背景差分相结合的车流量检测方法。该方法既解决了背景差分法对背景模型过于依赖的问题,也消除了帧差法对于快速运动的物体产生的双影问题。阴影问题一直是影响视频车流量检测准确率的重要因素,而RGB空间颜色分量有相关性。本文采用基于霍特林变换的阴影抑制,该方法通过帧差获得车辆和阴影的轮廓图像,然后对轮廓图像用霍特林变换解除了RGB分量的相关性,在此基础上构造了阴影测度进行阴影检测,具有较好的阴影检测效果。多车道车流量检测对车道的划分提出了更高的要求。Hough变换是一种能够有效地检测直线的方法,但要求图像清晰,车道标志线明显。本文对传统Hough变换进行改进,得到一种基于模糊理论的车道标志线检测新算法,该算法将模糊集和动态聚类分析的思想引入到Hough变换算法中,从而获得直线的精确定位,能适应不同环境下的车道划分。对处理后的图像选取局部检测区域并提取数据流,通过对数据流的校正、比较来实现车辆计数。更多还原

【Abstract】 The detection of the traffic flow is an important part in the Intelligent Traffic Control System. The traffic information can be got from the detection of the traffic flows in the real time. These information are used in developing the control signal to instruct the traffic which can reduce traffic j ams, enhance the using efficiency of the traffic facilities, and finally meet the requirements of the efficiency for the entire transport system. Based on video traffic detection has many advantages, such as wide detect regional, simple operation, comprehensive information extraction, so it is attracting more and more attention and becoming the emerging research hot spot.After analyzed and compared to the advantages and disadvantages of the traditional methods, a method that combines the inter-frame difference and background subtraction is offered in this paper. This method can resolve the problem of depending too much on the background in the method of background finite difference. At the same time, it also can solve the double problem of frame differential method for rapid movement of objects.The shadow is always an important factor in the accuracy of traffic detection. The variables in RGB space have correlation. In this paper, hotelling transform will be used to solve the problem of shadow. The outline image of the vehicles and shadow can be got through frame differential, and then transform the contour image by hotelling transform, which lifting the RGB component of the correlation. On the basis of the model for the shadow measure detection, the shadow can be removed successfully.Multi-lane traffic flow detection put forward higher request to lane line detection. Hough Transform can detect the lines effectively; however, it requires the clearer of images and also the driveway lane markings. In order to improve the traditional Hough Transform) this paper advances a new algorithm based on fuzzy theory in lanes detection. This algorithm introduces the fuzzy sets and dynamic clustering analysis to Hough transform algorithm, which can orient the lines accurately and mark off the driveways effectively in Varieties of environmental.Local detection area will be chose after image processed, and data flow will be extracted from the local image. the traffic flow can be detected after calibrating, comparison and calculating the data flow.更多还原