摘 要
近年来，随着我国国民经济的不断发展，交通运输越来越繁忙，对交通管理提出了新的要求。在过桥收费站、大型停车场、城市道路监管、治安卡口、港口和飞机场等实际交通系统中，需要对汽车车型进行识别，以便收取相应的费用和提高交通系统车辆监控和自动化程度。因此，如何对上述各类车辆收费站实现现代化的管理，具有重要的现实意义。
针对基于神经网络的汽车车型识别系统中的识别技术问题，本文从以下四个部分进行了研究和探讨：
第一部分，论述了车型识别的研究背景和意义，详细分析了我国目前车型识别系统的应用现状及研究现状，指出了目前国内外应用系统及其车型识别方法存在的缺陷与不足。
第二部分，提出了车型识别的模型。首先对采集的车辆图像进行预处理，通过灰度转换、图像平滑等方法剔除噪音，以提高图像质量。然后对其进行分割并提取特征，在这个过程中经过图像的二值化处理[16]，拉普拉斯边缘检测、图像横向填充与纵向填充、轮廓提取、图像修正，再提取出图像车型的上顶长、高、前底长、后底长等特征参数。结合所提取的特征参数进行车型识别。
第三部分，设计拉普拉斯边缘检测算子的汽车识别算法。采用序列差影法进行背景剔除；边缘检测之后的图像进行离散噪点的剔除。采用轮廓法对横/纵向填充图像进行轮廓提取。
第四部分，设计实现一个车型识别系统，以此检验论文理论研究的可行性，并通过不断地实际测验来改良算法。
本文以VC++6.0为软件平台，以Matlab为仿真平台，编程实现了基于序列图像的车型识别系统，通过实验数据分析表明，本文给出的识别方法能得到较好的识别结果。
关键词 ：智能交通，车型识别，图像处理，特征提取
I
ABSTRACT
With the development of our national economy, traffic has become heavier in recent years, which has put forward new requirements for traffic management. In actual traffic system such as bridge toll station, large parking lot, ports and airports, vehicle recognition is necessary to charge corresponding fees and improve the automation and vehicle monitoring of the traffic system. Therefore, the modern management of the toll stations above has important practical significance.
Facing the technical problems of the vehicle recognition system based on the neural network, this paper conducts research and discussion from the following four parts :
In the first part, we discuss the background and significance of the vehicle recognition, then detailed analyzes the present situation of the application of vehicle recognition system and the present research status, and point out the shortcomings of vehicle recognition systems at home and abroad.
In the second part, we put forward the model of the vehicle recognition system. First we conduct the preprocessing of the obtained vehicle image, to reduce the noise of the image by graying and smoothing it, to improve the quality of the image. Next step is the segmentation and feature extracting. After the image binarization processing, Laplace edge detection, image horizontal and vertical filling, contour extraction, image correction, we extract the length and height of the front and back of the car. Vehicle recognition can be done with the characteristic parameters extracted. In the third part, we design a vehicle recognition algorithm using Laplace edge detection operator, and eliminate background using sequence difference image method. Then eliminate the discrete noise points after edge detection of the image., and extract the contour of the horizontal/vertical filled image using contour method. In the fourth part, we designed a vehicle recognition system to test the feasibility of the theory, and improve the algorithm through the continuous actual test
II
Based on software platform of VC++6.0, with Matlab simulation platform, we finish the vehicle recognition system based on image sequences. Through the experimental data analysis showed, we can find that this method of identification presented can get good recognition results.
Key words：intelligent transportation, vehicle recognition, image processing, feature extraction
III
目 录
前 言............................................................... 1 1绪论 .............................................................. 3
1.1选题的目的及意义............................................. 3 1.2车型识别的研究现状........................................... 4 1.3论文研究的思路与流程......................................... 4 1.4车型图像处理平台............................................. 5
1.4.1车型图像硬件平台 ....................................... 5 1.4.2车型图像软件平台 ....................................... 5 1.4.3车型图像仿真平台 ....................................... 6
2车辆图像的分离与处理 .............................................. 8
2.1车辆图像的分离............................................... 8 2.2图像处理的基本知识.......................................... 11
2.2.1 BMP图像的读取 ........................................ 11 2.2.2图像的灰度化处理 ...................................... 11 2.2.3图像的平滑处理 ........................................ 12 2.2.4灰度图像的二值化处理 .................................. 13 2.2.5图像的边缘化检测 ...................................... 13 2.3车辆图像的处理结果.......................................... 16
2.3.1车辆BMP图像显示与读取 ................................ 16 2.3.2车辆图像灰度化 ........................................ 19 2.3.3车辆图像平滑处理——中值滤波 .......................... 20 2.3.4车辆图像的二值化及二值反向 ............................ 22 2.3.5车辆图像的边缘检测 .................................... 24
3车辆的特征提取 ................................................... 26
3.1特征提取.................................................... 26 3.2车型轮廓提取................................................ 26 3.3目标区域的填充.............................................. 29
IV
3.3.1图像横向填充 .......................................... 29 3.3.2图像的纵向填充 ........................................ 31
4车型识别 ......................................................... 32
4.1车辆轮廓识别................................................ 32 4.2车辆特征参数的提取与计算.................................... 32 4.3车型特征匹配................................................ 33 5总结与展望 ....................................................... 35
5.1总结与分析.................................................. 35 5.2未来展望.................................................... 35 致 谢.............................................................. 37 参考文献........................................................... 38
V