【信息技术】【2018】基于视觉特征的车辆自动检测与识别

本文为加拿大温莎大学（作者：Hao Lyu）的硕士论文，共84页。

近几十年来，汽车已成为世界上最受欢迎的交通工具。高精度和高成功率是车辆自动检测和识别的关键因素。车牌作为车辆上最重要的标签，是车辆的一种公共识别手段。但是，犯罪分子可以将其盗用并贴在不同的车辆上，以隐藏其身份。此外，在某些情况下，来自不同国家的两辆车的车牌号可能相同。在本论文中，我们提出一个新的车辆识别系统，提供高精度和高成功率。该系统分为四个阶段：车牌检测、车牌识别、车牌区域检测和车辆形状检测。该系统将特征转化为局部二值模式（LBP）和方向梯度直方图（HOG）作为训练数据集。为了在实时应用中达到较高的精度，提出了一种新的系统更新方法。同时，通过该系统，我们可以将车辆特征和信息存储在数据库中。此外，通过数据库比对，该程序可以自动检测出车牌和车辆之间的任何差异。

In recent decades, a vehicle has become themost popular transportation mechanism in the world. High accuracy and successrate are key factors in automatic vehicle detection and identification. As themost important label on vehicles, the license plate serves as a mean of public identificationfor them. However, it can be stolen and affixed to different vehicles bycriminals to conceal their identities. Furthermore, in some cases, the platenumbers can be the same for two vehicles coming from different countries. Inthis thesis, we propose a new vehicle identification system that provides highdegree of accuracy and success rates. The proposed system consists of fourstages: license plate detection, license plate recognition, license plateprovince detection and vehicle shape detection. In the proposed system, thefeatures are converted into local binary pattern (LBP) and histogram oforiented gradients (HOG) as training dataset. To reach high accuracy inreal-time application, a novel method is used to update the system. Meanwhile,via the proposed system, we can store the vehicles features and information inthe database. Additionally, with the database, the procedure can automaticallydetect any discrepancy between license plate and vehicles.

1. 引言
2. 文献回顾
3. 相关材料与方法
4. 结果与讨论
5. 结论

更多精彩文章请关注公众号：