基于遥感图像的城市道路及车辆信息采集研究

发布时间：2019-01-11 09:37

【摘要】：遥感技术的飞速发展，使得各种遥感图像分辨率不断提高，可携带信息也因此不断增加，因此使得从遥感图像下提取城市交通流成为可能。城市交通由车、路、人三要素组成，因此道路和车流信息成为城市交通的两个重要方面。为了准确从城市交通中提取这两个信息，本文以图像处理方法为基础，在充分分析遥感图像识别目标特征的基础上，重点围绕道路边缘提取，以及将提取道路边缘应用于城市交通流提取方面开展了研究，并提出了一种基于双阈值序贯相似性模板匹配的方法，，以及一种基于车道线内道路区域对车辆进行采集识别的交通流优化提取方法。主要研究内容如下： 1.针对遥感图像背景消除。通过对城市遥感图像的背景分析，分别采用多种掩膜方式实现了对植被树木、阴影部分的分离，以及应用基于GIS系统的多边形道路边缘逼近的方式，实现了对道路边缘的粗略提取。 2.在对图像道路边缘提取的方面，针对模板匹配效率低下，弹性空间小等不足之处，提出了一种基于双阈值序贯相似性模板匹配的方法，在提高了模板匹配速度的同时，对阈值的选取提供了弹性空间。并采用此方法实现了对一般道路边缘的提取，以及与其他常用边缘提取方法的比较。 3.针对一般车辆提取特征维度高，分类器过于复杂的等问题，提出了基于车道线内道路区域对车辆特征提取的优化方法，减少了车辆识别提取的特征数，降低了车辆分类器的复杂程度，提高了检测效率。 4.针对减少提取特征后的分类器，采用结合模板匹配的方式，实现对车辆的提取，进一步提高检测精度。实验结果表明，本文方法在不牺牲过多检测效率的同时，实现了对检测时间的大幅减低，因此可以视为提高了效率。通过详细论述了遥感图像下道路边缘以及交通流信息的提取流程以及关键技术，本文在最后对研究内容进行了总结并提出了下一步的研究方向。
[Abstract]:With the rapid development of remote sensing technology, the resolution of all kinds of remote sensing images has been improved, and the portable information has been increasing. Therefore, it is possible to extract urban traffic flow from remote sensing images. Urban traffic consists of three elements: vehicle, road and human, so road and traffic information become two important aspects of urban traffic. In order to extract these two information from urban traffic accurately, based on the image processing method, based on the full analysis of the target features of remote sensing image recognition, this paper focuses on road edge extraction. And the application of road edge extraction to urban traffic flow extraction is studied, and a method based on two-threshold sequential similarity template matching is proposed. And a traffic flow optimization extraction method based on the road area in the lane line to collect and identify vehicles. The main contents are as follows: 1. Background elimination for remote sensing images. Based on the background analysis of urban remote sensing images, the separation of vegetation trees and shadow parts and the approach of polygonal road edge approximation based on GIS system are realized by using a variety of mask methods, respectively. The rough extraction of road edge is realized. 2. In the aspect of image road edge extraction, aiming at the inefficiency of template matching and the small elastic space, a method of template matching based on two-threshold sequential similarity template matching is proposed, which not only improves the speed of template matching, but also improves the speed of template matching. The elastic space is provided for the selection of threshold. The method is used to extract the general road edge, and to compare it with other commonly used methods. 3. In order to solve the problems such as the high dimension of vehicle feature extraction and the complexity of classifier, this paper proposes an optimization method for vehicle feature extraction based on road area in lane line, which reduces the number of features extracted by vehicle recognition. The complexity of vehicle classifier is reduced and the detection efficiency is improved. 4. Aiming at the classifier after reducing the feature extraction, the method of combining template matching is adopted to realize the vehicle extraction, and the detection accuracy is further improved. The experimental results show that the proposed method can greatly reduce the detection time without sacrificing too much detection efficiency, so it can be regarded as improving the efficiency. In this paper, the extraction process and key technology of road edge and traffic flow information under remote sensing image are discussed in detail. At the end of this paper, the research content is summarized and the next research direction is put forward.
【学位授予单位】：湘潭大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP751

【共引文献】