Study And Application Of Transportation Information Extraction Based On Remote Sensing Images

High-resolution remote sensing images embody abundant transportation data, including the road network, the number and distribution of traffic and so on, which makes the images good transportation information sources. In this paper, the current extensive used QuickBird satellite images are applied, with which a series of transportation information extraction experiments are performed. Based on the experiments, this paper proposed a series of traffic information extraction processes and methods.This paper starts the research work through the analysis of the characteristics of both the remote sensing images themselves and the transportation information embodied in them. The thesis sums up the characteristics of high-resolution remote sensing images, especially the road network and traffic modalities. The modalities of the road network are analyzed as two sorts of objects, which are road links and intersections. The modalities of the traffic include the form of traffic queues and individual vehicles.On the other hand, the paper summarizes the current important image processing methods adapted to high-resolution remote sensing images, and different color models are compared. Based on the result of the contrast, the HSI color model is applied as the practical one for the follow-up study of this paper.On this basis, in-depth researches of road network and traffic information extractions are carried out, and experiments are performed to validate the feasibility of the methods. In the study of road network extraction, the thesis firstly performs the road links extraction, and proposes three technologies including grey morphology, texture analysis and the method based on color information, the first two methods of which do not concern about any color characteristics. After that the pros and cons of the three methods are compared and analyzed. On the basis of road links extraction, the paper presents road intersection extraction methods for both plane intersection and three-dimensional interchange, and also the corresponding experiments are carried out. In the study of traffic information extraction, the thesis proposes two methods for individual vehicle extraction including the morphological technology which does not consider any color characteristics and the color information based vehicle extraction process, and develops some new methods such as vehicle color selection technology and moving window counting method. Finally, on the basis of single vehicle extraction, this paper studies the applications of traffic information, such as road section traffic density measurement and the calculation of the queue length.The experimental results indicate that the various methods presented in this paper are certain extent effective. The contrast analysis of some methods in this paper shows that certain methods are applicable for different transportation information extraction. Operators need to choose the suitable method for given transportation information type according to the actual instance.The results of this paper are helpful to promote the application of the remote sensing industry achievements in the field of transport, as well as improving the degree of information for the transport field.