【作者】 贾杰；

【导师】 冀小平；

【作者基本信息】 太原理工大学 ， 电子与通信， 2012， 硕士

【摘要】 本文以视频道路交通监测系统中运动车辆的检测与跟踪为应用背景,对运动目标检测中背景更新、阴影消除、运动目标跟踪和摄像机自动定标等一些难点问题进行了深入的分析研究,提出了一些新的算法,主要内容如下：1.分析了目前运动目标分割和检测的主要方法。包括运动优先的算法和分割优先的算法。提出了一种新的背景重建算法。算法以图像中任何一像素点背景出现概率最大为准则。特点是节省内存,跟新速度快,统计帧数多。实验验证可靠有效。2.针对阴影干扰的消除,提出一种结合灰度信息和结构信息的阴影干扰消除算法,算法首先进行阴影类型判别,在此基础上,用统计的方法检测目标物与阴影的分界点,并用多个分界点中拟合一条分界线作为目标物与阴影的分割分界。取得较好的实验效果。3.分析了目前运动目标跟踪的主要方法。提出一种结合面积校正的改进的目标跟踪算法。因为成像系统的原因,同一目标物在不同帧中处在不同的位置,面积也大小不同。给运动目标跟踪带来一定误差。本算法弥补了这一误差,结合用CM(1,1)模型进行预测。取得了较好的实验效果。4.分析了摄像机定标的主要方法。在智能交通中的摄像机定标,可利用的先验知识是图像序列中的运动目标物是车辆,而其中占多数的是小轿车。假设小轿车的实际面积已知。本文提出一种利用这些先验知识的摄像机自动定标算法。实验效果可靠有效。本文所提出的运动车辆检测、跟踪方法、摄像机自动定标算法和阴影干扰消除算法作为视觉监控领域的一种普遍方法,具有一定的理论意义和实用价值,可以推广到视觉监控的其他应用中,具有广阔的应用前景。更多还原

【Abstract】 In this paper, the detection and tracking of moving vehicles in the visual road traffic detection system as application background, conducted in-depth analysis and study of the problem of background update, shadow elimination, moving object tracking and camera auto-calibration on moving target detection, proposed some new algorithms. The main content is as follows:1. This paper analyzes of the current moving object segmentation and detection methods. Including the movement of priority algorithms and split the priority algorithm. Put forward a new algorithm of background reconstruction. Algorithm to image any pixel background probability criterion. Characteristic is to save memory, with the new speed, count the number of frames much. Experimental validation is reliable and effective.2. For the elimination of shadow interference, puts forward a shadow interference cancellation algorithm that combination of gray information and the structure information, for shadow type discrimination, on this basis, using statistical methods to detect the target and shadow boundary, and using multiple cutoff point fitting a line as a target and shadow segmentation boundary, in orders to better experimental results.3. Through the analysis of current target tracking method, puts forward a improved target tracking algorithm that combination area correction. Because of the imaging system, the same target gives some error of moving target tracking in different frames in a different location, different area size. The algorithm makes up for this error, combining with CM (1,1) model to forecast, and obtained a better experimental effect.4. It analyzes the main methods of camera calibration. In the intelligent transportation of camera calibration, the available prior knowledge is of moving object in image sequence is the vehicle, which account for the majority of the cars. If the actual area of the car is known, this paper puts forward a CCD automatic calibration algorithm of using these priors knowledge. The experimental result is reliable and effective.The proposed algorithms of moving vehicle detection, tracking, camera automatic calibration and shadow interference cancellation, have certain theory significance and practical value, can be extended to other applications in visual surveillance, and have broad application prospects.更多还原