基于卡口图像的车型识别方法研究

发布时间：2018-06-16 22:26

本文选题：卡口图像采集系统 + 车型识别　；参考：《合肥工业大学》2015年硕士论文

【摘要】：作为智能交通系统的一个重要部分,车型识别在打击盗窃车辆、规范交通秩序、高速公路自动计费等方面具有广泛的应用前景。车辆种类繁多,结构相似、车标类似的车辆大量存在,同时受到复杂背景因素(如：光照、天气)及一些外在因素(如人为污损、无牌照)的影响,获取车辆的准确信息非常困难。因此,车型识别成为计算机视觉领域的一大研究热点和难点。本文以安徽省科技攻关项目“基于视频图像的智能车型自动识别系统研究”为背景,针对基于卡口图像的车型识别技术进行研究。通过首先进行车前脸定位,然后基于特征提取和识别技术,进行车标和车前脸的识别,以此来进行车型的判断。本文的主要工作包括以下几个方面：1)车前脸的定位：在对传统车牌定位方法进行了解的基础上,提出了基于多特征融合的车牌定位方法,该方法基于颜色特征和纹理特征,对车牌进行定位。在定位好车牌的基础上,基于边缘检测和车牌位置,实现了车前脸的定位。2)基于车标和车前脸的车型识别：针对车前脸的特征提取,使用常见的特征提取方法训练分类器实现车型识别,其准确率不能满足要求,因此本文提出通过卷积神经网络自动学习车前脸的特征,以实现对车辆的识别。考虑到车辆类型的复杂性,首先进行车标识别以辅助系统判断车辆的品牌,然后使用车前脸的特征模型识别车辆系列。3)车型识别系统的实现：基于前面两章中提到的技术,开发了基于卡口图像的车型识别系统,基于测试数据集,对该系统的效果进行了评估和分析。
[Abstract]:As an important part of Intelligent Transportation system (its), vehicle identification has a wide application prospect in anti-theft of vehicles, regulation of traffic order, automatic charging of freeway and so on. There are a wide variety of vehicles with similar structures and similar markings, and they are also affected by complex background factors (such as light, weather) and some external factors (such as man-made defacement, unlicensed). It is very difficult to get accurate information about the vehicle. Therefore, vehicle recognition has become a hot and difficult point in the field of computer vision. In this paper, the technology of vehicle recognition based on bayonet image is studied based on the research of intelligent vehicle automatic recognition system based on video image, which is the key project of Anhui province. By first locating the front face of the vehicle, then based on the feature extraction and recognition technology, the identification of the vehicle logo and the front face of the vehicle is carried out, so as to judge the vehicle model. The main work of this paper includes the following aspects: on the basis of understanding the traditional license plate location methods, a license plate location method based on multi-feature fusion is proposed, which is based on color features and texture features. Position the license plate. On the basis of locating license plate, based on edge detection and license plate location, the vehicle front face location. 2) vehicle identification based on vehicle sign and front face: for vehicle front face feature extraction, The common feature extraction method is used to train the classifier to realize the vehicle recognition, and its accuracy can not meet the requirements. Therefore, this paper proposes to automatically learn the features of the front face of the vehicle by convolution neural network in order to realize the recognition of the vehicle. Considering the complexity of vehicle type, we first identify the vehicle logo to assist the system to judge the brand of the vehicle, and then use the feature model of the front face to recognize the vehicle series. 3) the realization of the vehicle recognition system: based on the technology mentioned in the previous two chapters, The vehicle recognition system based on the bayonet image is developed. Based on the test data set, the effect of the system is evaluated and analyzed.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495

【相似文献】