【信息技术】【2015.05】一种车辆检测与跟踪的图像处理算法

本文为印度Rourkela国立技术学院（作者：Soumyashree Seth）的计算机科学与工程硕士论文，共52页。

这里写图片描述

目标检测、跟踪及其速度估计在交通监控系统中起着重要的作用，它是图像处理领域中一种新兴的应用技术。实时对象跟踪是许多计算机视觉实际应用中的重要工作。在典型的视觉跟踪器中可以分为两个主要部件。从大多数运动相机框架访问的图片来确定特征是不可靠的，作为遵循多种跟踪目标的相同规则，虽然它们之间具有相似性，但对于不同目标的识别并不是那么简单。

在本文中，我们描绘了一个用于跟踪车辆的系统，运动图像中的车辆数量是模糊的，不同的帧之间也是不同的。我们在图像中识别出所有运动物体，并且为了跟踪车辆，我们利用卡尔曼滤波器和阴影高亮算法将其从不同的运动背景中分离出来。因此，该系统可以完全识别和跟踪所有车辆。本文提出的算法可以适用于许多运动目标的问题。该算法通过利用卡尔曼滤波器对小范围内目标的位置、形状和大小进行估计，从而得到目标的运动轨迹。为了突出可见的运动目标，需要去除图像中通过形态学运算产生的噪声。通过图像序列中的特征匹配来跟踪不同的目标，匹配的对象具有相同的编号，最后通过计算图像中存在的车辆速度来估计该片交通区域速度中的平均车速。如果平均速度满足设定的某个阈值，那就意味着交通拥挤。在整个处理的过程中，CCTV摄像机被认为是静止的。

Object detection ,tracking and its velocity estimation play importantroles in traffic surveillance system.It’s an arising and On-Demand technique inthe field of image processing . Real time object track is vital job in manycomputer vision practical application.Two major components can be distinguishedin a typical visual tracker.The picture determination of the feature accessiblefrom most movement camera framework is low. As a rule for following multi item,recognizing them from another isn’t simple due to their comparability. In thispaper we depict a system, for tracking various items, where the articles arevehicles. The quantity of vehicles is obscure and differs from frame to frame.We identify all moving objects, and for tracking of vehicle we utilize thekalman filter and shading highlight and separation of it from one casing to thefollowing. So the system can recognize and follow all vehicles exclusively. Theproposed method can be utilized in numerous moving items. The proposedalgorithm uses kalman filter to estimate objects’ location,shape and size in asmall range and to gain trajectory a of the object. To make visible movingobject prominently it is required to remove the noise in the image which isdone by morphological operations.Objects are tracked by feature matching inimage sequence.Matched objects are numbered same .Then number of objectscrossing through a particular line in the image are counted. Finally to figureout rush in traffic area velocity is estimated by averaging the speed ofvehicles present in a frame. If average velocity accedes a threshold value,thatmeans heavy traffic is there. Through out the process CCTV camera is consideredstationary.

这里写图片描述