【作者】 杜宇；

【导师】 毛佳；

【作者基本信息】 吉林大学 ， 物流工程， 2014， 硕士

【摘要】 在当今社会，各行各业的竞争越来越大，良好的服务体系越来越引起企业的重视，很多企业在这一方面投入了大量的财力、物力，以求能提高自身的服务水平。这一特点在物流行业更为突出，物流行业的服务性能十分突出的，它通过运输客户物品来实现其良好的服务性能。而物品在运输过程中，如何实现透明化，如何更好的调动车辆，已经成为至关重要的问题。构建一个有效的车型识别系统就能有效的解决这一突出问题，但是目前车辆监控体系的价格十分昂贵，这让很多的物流企业感到力不从心，特别是对于一些中小型的物流企业来说，他们更多关注的是投放出去的成本是否能收获利益。因此，如果有一种较为经济、实用的动态监控系统，则必将为中小型物流企业带来许多便利之处。与此同时，在道路交通体系中，如果能够设计出专门针对物流车辆的车型识别系统进行监管、收费等，则对物流企业来说，无疑是对运输服务巨大的提高。本论文将微分几何方法运用到车型识别当中，具体来讲，先应用计算机系统对过往物流车辆进行图像的获取、轮廓的分割，之后应用本文所建立模型计算出轮廓上所有特征点的曲率和相似距离，并将所有物流车型轮廓上点的相似距离和曲率存储在数据库中，当识别系统运用时，把待匹配车型的相关参数与数据库中数据进行匹配，匹配成功，则输出车型型号，即达到了识别目的。论文的大体研究思路如下：首先介绍了研究物流车辆识别系统的意义、国内外专家学者的研究现状以及本论文所用到的研究方法等；接下来是相关理论知识的搜集和研究，主要介绍微分几何方法的产生和发展，通过研究微分几何的发展历程，对微分几何方法有了一个很清晰的认识。并讨论了现有车型识别系统的识别原理和微分几何方法运用到车型识别系统中的可能性；并且应用所采集到的物流车型图像通过对图片进行灰度化、去噪、轮廓提取等步骤，抽象出车辆的外观轮廓图像。之后为物流车辆的类型分类、车辆轮廓的提取步骤、特征点的描述及存储、以及模型的对比（本论文主要比较微分几何KMP模型和BP神经网络模型）等，通过这一系列的研究，建立了一套完善的物流车型识别系统模型，之后我们又根据模型本身存在的不足，论文又研究了在特殊情况下该识别系统的完善方法，使其满足了一般性要求。为了使得本系统更加的完善，本文随后建立了物流车型的数据库，用于存储车辆外观轮廓上特征点的曲率以及相似距离。最后指出了本模型的创新点和不足之处，为后续研究提供了方向。更多还原

【Abstract】 Nowadays, more and more industries take good service as the most significantcontent of management. As a result, various industries have invested a lot of itsmanpower and resources. Those industries strive for the further improvement of theirservice levels, and an enhancement of their overall competitiveness. Thisphenomenon happens mainly in the logistics industry. Logistics industry is aspecialized service sector; it achieves its goal of service through the vehicle transport.“How to achieve transparency in the transport process and meet the requirements ofcustomer” has become a key issue of logistics industry. In face, the effective way tosolve this problem is to build an effective vehicle recognition system. But the currentprice of the vehicle recognition system is very expensive, which makes a lot oflogistics enterprises feel helpless. Especially for those small or medium sized logisticsenterprises, they are more concerned with the benefits of their cost. Therefore, if thereis an economical vehicle recognition system, it certainly will bring a lot ofconvenience to small and medium sized logistics enterprises. Meanwhile, if we candesign a vehicle recognition system which is specifically for the logistics vehicle, itwill bring a huge improvement for the transport services of logistics enterprisesundoubtedly.This thesis will apply the differential geometry to identify the types of vehicles.Specifically, first, we use computer systems to accomplish the task of logisticsvehicles’ image acquisition and contour segmentation. Then, we will use the modelthat established by the writer to calculate the curvature of all characteristic points andsimilarity distance. Afterwards, we will store the datum in the database. When we runthe identification system, we will the matched model parameters with the data in thedatabase. If the matching is successful, it will analyze the type of the vehicle, and wecan achieve the goal of vehicle recognition.The general idea of the paper: The first chapter is the introduction. It mainly introduces the significance of the research of the vehicle identification system,research status at home and abroad and the search methods. The second chapter ismainly talks about literary review of relevant theories. It focuses on the emergenceand development of differential geometric; existing principle of vehicle recognitionsystem; the possibilities of apply differential geometry in vehicle recognition system.The third chapter focuses on the extraction of vehicle profile and the classification oflogistics models. This chapter includes the acquisition of vehicle image, grayprocessing, enhancement of gray processing, extraction of contour and otherprocesses. The fourth chapter is mainly about the storage of each model parameter,matching method and the establishment of vehicle recognition system. In this paper,the primary model parameter is the curvature and the relative distance between eachtwo points. Recognition model using BP neural network model to conductidentification; and demonstrates the complexity and robustness of this approach. Thefifth chapter is mainly about the establishment a database. This database is used tostore the data of vehicle types. The sixth chapter is the summary outlook of this thesis.It analyzes the innovation and deficiencies of this paper, and points out the subsequentresearch direction.更多还原