基于bp神经网络的车型识别_毕业设计论文.docx

基于BP神经网络的车型识别 摘要 车型分类识别技术作为智能交通系统中的关键技术，对提高道路运输效率，改善车 辆收费检测等方面有着重要的理论与现实意义。本文基于视频检测技术，首先通过图像 预处理、车辆分割、轮廓提取得到车辆的轮廓图，从中获得车辆的外形几何参数，并做 相关性分析，提取特征向量。然后利用提取的特征向量，构建BP神经网络的车型分类 系统进行车型识别。主要研究内容包括：(1)车辆检测研究。本文采用一种基于背景差分 的车辆分割方法，较好地解决了复杂交通情况下车辆的检测问题。(2)车型特征提取。根 据车型分类的需要，分析了车型特征参数的选择问题，为车辆分类奠定了基础。本文最 终选取了顶长比、顶高比、前后比作为特征向量。(3)车型分类研究。研究了基于BP神 经网络的车型分类，通过选择合适的特征参数，获得了较高的分类正确率。 应用效果与仿真结果表明，基于BP网络的车型分类技术的实时性、精确性和分类 识别性能等关键指标得到明显的改善，达到系统设计的预期要求。同时，我们采取的方 法具有提取的特征简单、量少，并且所构成的具有分类功能的BP网络简单、便于硬件 实现、有利于 BP 网络的分类识别等优点。 关键词： 智能交通系统；车型识别；车辆检测；特征提取；BP神经网络 Vehicle Recognition Based on BP Neural Network Abstract As the key technology of Intelligent Transportation System(ITS),vehicle recognition has all important theoretical and practical significance in improving the efficiency of road transportation and testing of vehicle charging.Firstly, the paper based on video detection discusses how to get the vehicle contour map through these operations such as image pre-processing,vehicle segmentation and contour extraction to derive geometrical parameters of vehicles which are used to establish the vector by a correlation analysis.Secondly, we use these feature vectors to build the system of vehicle classification based on BP Neural Network to recognize the vehicles.The main tasks are as follows:(1)Vehicle detection. This paper presents the vehicle segmentation method based on background subtraction. It can solve the problem of vehicle detection in complex traffic situations effectively.(2)Feature extraction.According to the needs of vehicle classification, we analyze the selective problems of the parameters to laid the foundation of vehicle classification.This paper finally adopts the vectors of the ratio of top and length,top and height, forward and back.(3)Vehicle classification.This paper studies vehicle classification based on BP neural network and obtains higher classification accuracy by selecting the appropriate par