【作者】 孙文彬；

【导师】 岑敏仪；

【作者基本信息】 西南交通大学 ， 大地测量学与测量工程， 2003， 硕士

【摘要】 随着交通科技的发展，为了解决日益增长的路网资源需求和缓慢的建设速度之间的矛盾，人们提出了智能交通系统的概念，利用智能化的交通管理来挖掘路网资源的最大潜力。导航定位是智能交通系统中的一个核心问题。而在城市交通管理的应用中，由于卫星定位信号经常会受到建筑物的遮挡和干扰，卫星定位无法满足导航的实际要求，通常需用地图匹配和航位推算器(Dead-Reckoning，DR)的方法进行辅助导航，这种导航方式称为组合导航系统，其中地图匹配可用来修正DR的方向和距离的系统误差，以便在卫星信号失锁时，提高DR自主定位的精度。 地图匹配的核心是匹配技术。匹配技术是计算机技术、模式识别、人工智能等技术的融合。匹配技术在摄影测量、导航定位、人工智能识别、计算机视觉、机械制造、医学、遥感、GIS、绘图、文物考古、环境保护等领域都有着广泛的应用。为此，匹配技术吸引着众多的中外学者的关注，也是当今测绘界的热点研究课题之一。 本文着重研究采用曲线匹配方法修正DR的距离系统误差。Woleson等人提出的曲线匹配算法，在很大程度上依赖于特征点的提取和特征点的对应。特征点提取的准确度和对应直接关系着匹配的精度，甚至关系着匹配的成败。在实际情况中，特征点的提取存在着一定的误差，这必然会加大对应特征点寻找的难度，甚至在某些情况下会提出错误的对应关系。本文在Woleson等人的研究基础上，提出用遍历的方法来减小对特征点的依赖度。文中首先假设曲线中的每个节点都可能是待匹配曲线的起点，然后根据等长度分段，比较两匹配曲线之间的误差绝对值，由此判断曲线是否相似。针对DR导航轨迹的具体情况，在Woleson的匹配算法基础上，对匹配过程进行了改进，使其更能适合实际情况。最后，利用香港地区的实际导航数据，分别采用改进算法和Woleson算法进行计算，分析试验结果，证明改进后的算法不但降低了对特征点的依赖，而且在匹配中充分利用了曲线特征信息，简化了曲线匹配算法。但是，改进的算法反映最大间隙差(误差) 西南交通大学硕士研究生学位论文第11页的灵敏度稍弱，这对曲线特征不明显(如近似直线)的线段，算法的效果不明显。更多还原

【Abstract】 With the development of transportation, a concept of intelligent transportation systems is put forward to resolve the confliction between requirement of increasing road-network resource and their constructing speed. The potential of road-network resource is excavated by intelligent management. Navigation is one of the most important parts of intelligent transportation systems. In the management of city transportation, navigation of satellite cannot meet the practical requirement because of shielding and interfering of satellite signal. Therefore, map matching and dead-reckoning (DR) is used to assistant navigation. This combination is defined as GPS/DR integrated navigation system, in which the map matching is used for correcting the orientation and distance errors of DR. The matching method will improve the navigational precision based on DR, when GPS and/or GLONASS signals are lost.Matching technology is kernel point of map matching. It is integrated with computer, pattern recognition, and artificial intelligence. Matching technology is widely used in photogrammetry, navigation, artificial intelligence, computer vision, machine manufacture, physic, remote sense, GIS, drawing, archaeology, environmental protection and so on. Therefore, many scholars are attracted by matching technology. Nowadays, this technology is also one of research hotspots in surveying and mapping fields.This article pays attention to make use of curve matching to correct the distance error of dead-reckoning. Matching algorithm improved by Woleson is dependent on redrawing characteristic point and corresponding relation. The accuracy of redrawing characteristic point and finding corresponding relation determines the accuracy of matching, even success of matching or not. In fact, there are some errors in redrawing characteristic point. Therefore, this is inevitable to increasedifficulty to find the corresponding relation between characteristic points. In some case, the improper corresponding relation is found by Woleson’s algorithm. Based on the study of Woleson, a way of point traversal is put forward in this article. The method can decrease the number of finding the characteristic points. It is firstly suppose that each node may correspondingly match curve’s original point. Secondly, the curve to match is divide into some sections. The absolute errors between two matching curves are compare in every section. Then, the resemblance of two curves is judged by their absolute errors. Based on Woleson’s matching algorithm, matching process is mended. The new algorithm is more suitable for practical application in matching between vehicle navigation track based on DR and electronic map than old one. At last,some practical navigation data in Hong Kong are processed by algorithm improved and Woleson’s. It is proved by the results of experiments that new algorithm not only decreases the number of characteristic point, but also predigests curve matching algorithm. The new algorithm makes full use of curve characteristic information in matching process. On the other side, the new algorithm is not sensitive to max errors between two matching curves. When curve characteristics are not obvious (as line), the matching result is not ideal.更多还原