【信息技术】【2014.07】交通监控中的车辆跟踪与速度估计

在这里插入图片描述
本文为加拿大滑铁卢大学（作者：KURTIS NORMAN MCBRIDE）的硕士论文，共80页。

车辆跟踪是目标检测和跟踪的关键应用之一。交通监控在如今这个时代已经变得至关重要，目前，道路上的车辆数量已经大大增加。为了保护驾驶员的安全，交通执法部门在全国各地的路段都指定了限速。然而，不负责任的司机仍然会超过限速，因为他们知道自己不太可能被抓到并接收惩罚。本文开发了一种能够在视频中检测运动车辆并显示行驶速度的系统。如果车辆超过允许的速度限制，车辆实际速度将显示在视频中的车辆旁边，以便交通执法人员能够根据显示的速度采取必要的措施。该系统以Matlab/Simulink为仿真平台，为阈值、滤波和Blob分析提供了综合的工具。光流（optical flow）是用来确定运动车辆的图像处理技术，采用中值滤波方法去除阈值图像中的椒盐噪声。多种形态学技术的组合被用来矫正处理后的图像，Blob分析在运动目标周围生成矩形框。矩形框的质心用于确定每辆车在给定框架上的位置。为了弥补深度感知信息的缺失，摄像头的高度和与道路之间的夹角是固定的，因此可以确定车辆接近摄像头的速度。结果表明，该系统能够成功地检测到车辆并显示其速度，且显示的速度误差相对较小。显示速度设置为每两帧更改一次，以便更容易进行观察。

Vehicle tracking is one of the critical applications of object detection and tracking. Traffic surveillance has become crucial in this day and age where the number of vehicles on the road has risen considerably. To preserve the safety of motorists, traffic law enforcement assign speed limits at different locations throughout the country. However, irresponsible motorists still exceed the speed limit since they know it is unlikely that they will get caught. In this paper, a system is developed which is capable of detecting moving vehicles in a video and display the vehicles speed as it goes. Should a vehicle exceed the allowed speed limit, it will be displayed in the video alongside the vehicle so that traffic law enforcers will be able to take necessary action based on the displayed speed. The system uses Matlab/Simulink as a simulation platform as it provides comprehensive tools for thresholding, filtering and blob analysis. Optical flow was the image processing technique used to determine the moving vehicles. A median filter was used to remove salt and pepper noise from the thresholded image. Combinations of several morphological operations were used to rectify whatever that is left. Blob analysis produces rectangles around the moving objects. The centroid of the rectangle is used to determine the location of each vehicle at a given frame. To make up for the absence of depth perception, the camera’s height and angle from the road is fixed so that the rate of which a vehicle approaches the camera can be determined. The results show that the system successfully detects vehicles and displays its speed, though there is a relatively small margin of error for the displayed speed. The displayed speed is set to only change once every couple of frames so that it would be easier to see.

1 引言

1.1 车辆跟踪与速度估计简介

1.2 问题描述

1.3 研究目标

1.4 研究内容

2 相关文献回顾

2.1 案例研究

2.2 噪声滤波

2.3 图像分割

2.4 闭合（形态学）

3 研究方法

3.1 整个项目流程图