基于深度学习的车牌识别方法研究

论文总字数：21812字

目 录

1绪论 8

1.1 课题背景 8

1.2 系统设计的目的和意义 8

2. 相关研究与技术 9

2.1 国内外研究现状 9

2.2 卷积神经网络 9

2.2.1 卷积神经网络简介 9

2.2.2 卷积神经网络的层级结构 10

2.2.3 卷积神经网络的训练过程 13

2.2.4 卷积神经网络的优点 14

2.3 计算机图像处理技术 15

2.3.1 概述 15

2.3.2 图像的分类 15

2.3.3 图像处理的常用方法 15

2.4开发工具介绍 16

2.4.1 Python简介 16

2.4.2 Tensorflow介绍 17

2.4.3 Linux介绍 17

2.4.4 opencv介绍 17

2.4.5 Keras介绍 18

3 系统的设计与实现 20

3.1 系统开发环境 20

3.1.1 硬件环境 20

3.1.2 软件环境 20

3.2 系统整体框架 20

3.3 图像预处理 21

3.3.1 色彩图像灰度化 21

3.3.2 图像的滤波 22

3.3.3 二值化 22

3.3.4 边缘检测 23

3.4 车牌定位 24

3.5 车牌字符分割 24

3.6 车牌字符识别 26

3.6.1 识别方法 26

3.6.2 神经网络训练与识别率 28

4 系统运行测试 30

总 结 32

参考文献 33

致谢 34

基于Python的车牌识别系统的设计与实现

章越

，China，zhangyue

Abstract: With the continuous development of the times, cars have entered more and more people's lives, and the management requirements for license plates are also getting higher and higher, so the license plate recognition system has solved this problem very well. The license plate recognition technology is a research hotspot in recent years and it is one of the results of applying image recognition technology to life. A complete license plate recognition system mainly consists of three technical steps: positioning of license plates, character segmentation of license plates, and identification of license plate characters. Since the recognition of license plates has become a major hotspot, many scholars have published many effective technologies and methods for their exploration and research. Based on the domestic license plate model and characteristics, this article reviews the outstanding achievements of deep learning in recent years and uses the convolutional neural network model proposed by alex in 2012 to complete a high-precision and speed license plate recognition system.

There is also a corresponding problem in the license plate recognition industry. Many license plate recognition is still an old management method. Lack of comparative management tools and means of processing data does not integrate all information, information, and can not keep up with the pace of the times. Based on this major premise. In order to fit the social status quo, we chose the topic of license plate recognition as the topic of this graduation project. The graduation project starts from a practical point of view and structures the license plate recognition mode under the new situation.

This graduation project uses Python as a programming technique to implement the overall framework of the software. The core technology of this system is constructed using a series of deep convolutional neural networks (CNN) interleaved with precise and effective algorithms. From the overall structure to solve the difficult problem of license plate recognition. Through the implementation of deep learning techniques, the overall scheme is designed from large modules, and the small modules refine the actual functions. After many analysis and verification, this system has strong practical value and development potential. At the same time with high accuracy, easy management and so on.

剩余内容已隐藏，请支付后下载全文，论文总字数：21812字