Research On A Algorithm For Dynamic Generation Of Vector Road Map Based On Bus GPS Traces Data

As an important infrastructure of intelligent transportation systems, the network map of roads plays a key role in many applications. Because of services provided by those applications depends heavily on network map of roads, which assures the high precision and accuracy. Otherwise the information acquired from the network map of roads may contain errors, which will bring great inconvenience for users.Traditionally, the method for generation of the road network map can be classified into two categories, one is surveying and the other is extraction from digital images. However both of the method we mentioned above are of long cycle and high cost, what's more, the data source is often out of date, which makes the results may not be consistent with the reality. With the significant improvement of the accuracy of GPS and the development of the small GPS receivers, there are a lot of vehicles have installed navigation system based on mature GPS technology, making collect vast traffic data possible. There are abundant geo-spatial information exist in GPS data, such as road structure, position of traffic signals,.etc. Because GPS data have advantages of good real time ability, high precision, wide origin,.etc, it has became a new data source used to generate the road network map. The routes of the bus cover almost all the main road of a city and its GPS data are easily available, so this paper presents a method for generating vector road network map of the city dynamically. There are three steps in this algorithm. Firstly, on the basis of in-depth analysis of the running bus features, errors and redundant data are eliminated, after that, seed points are filtered out based on the principle of statistics, which can represent the prototype of the road. Secondly, location of the seed point are corrected and clustered, after that, feature points that can determine the geography of the road are calculated out. Then the feature points are ordered along the road direction and smoothed, after that. Finally, the whole road can be represented by a sequence of ordered vector edge. In the last step, we will use a single road as the unit of input, dynamically detect and update topology structure of the existing road network. This step has two operations, one is cross detection and the other one is repeat sections recognition.After the network map of roads is generated, by using the formatted data interface for programmers provided by Google Earth, we display the topology structure of the roads intuitively. The experiment results proved that the method presented by this paper features high accuracy and feasibility, which able to dynamically generating the vector road network map in a short period of time at a very small cost.