本文为美国密苏里大学（作者：XINLI）的硕士论文，共69页。
本文利用支持向量机（SVM）分类器结合HOG（Histogram of Oriented Gradients）特征，提出了一种车牌检测与识别方法。该系统在不同的尺度上进行窗口搜索，使用支持向量机分析HOG特征，并使用均值漂移方法定位它们的边界框。提出了一种车头和车尾的检测方法，以加快扫描过程。对不同单元大小和块大小的HOG特征进行了性能比较，并进行了四轮拔靴法（bootstrapping），以获得更好的检测性能。我们的车牌检测结果表明，该方法对光照变化、车牌图案、摄像机视角和背景变化相对不敏感。我们在Caltech数据集（1999）上进行了测试，达到了96.0%的检测率。我们还研究了不同程度的噪声和运动模糊对其性能的影响。
在车牌检测之后，我们使用具有HOG特征的SVM分类器进行字符分割和识别。在字符分割中，我们需要处理低对比度和倾斜的平面。该系统在不同尺度下进行窗口搜索，利用支持向量机对HOG特征进行分析，并使用均值偏移定位其边界框。
In this work, we develop a license platedetection and recognition method using a SVM (Support Vector Machine)classifier with HOG (Histogram of Oriented Gradients) features. The systemperforms window searching at different scales and analyzes the HOG featureusing a SVM and locates their bounding boxes using a Mean Shift method. A carhead and rear detection method was also proposed to accelerate the timeconsuming scanning process. A comparison of the performance for different celland block sizes of HOG feature is provided, and four rounds of bootstrappingwas performed to achieve better detection performance. Our license platedetection results show that this method is relatively insensitive to variationsin illumination, license plate patterns, camera perspective and backgroundvariations. We tested our method on the Caltech data set (1999), and achieved adetection rate of 96.0%. We also studied how its performance is impacted bydifferent levels of noise and motion blur. After license plate detection, weproceed to perform character segmentation and recognition using SVM classifierswith HOG features. In character segmentation, we need to deal with low contrastand tilted plates. The system performs window searching in different scales andanalyzes the HOG feature using a SVM and locates their bounding boxes usingMean Shift.
1 引言与项目背景
2 基于HOG特征的车牌检测
3 车牌识别
4 结论与未来研究方向
下载英文原文地址：
http://page2.dfpan.com/fs/6lcje221c291764e472/
更多精彩文章请关注微信号：