| After nearly half a century of development, GIS technology, which was first proposed in the 1960s of last century, has been very mature so far. And with the continuous improvement of computer technology, digital maps of various scales have also been drawn, and in the past 20 years in our country, medium and small scale GIS database has been established. The important issue the current geographic information database currently facing is to update it so that reflecting the current situation of the geographical distribution. There are many means for its update, the conventional artificial revision of mapping is the main way of the present, and it is of course time consuming, high cost and heavy workload. With the development of Remote Sensing and Photogrammetry technology, the earth-observed images become more and more, and their spatial resolution is higher and higher, therefore, extraction of surface features from the image for the spatial database updates will greatly reduce the manual workload, and it is of great importance. It is also an important means of combination for the Remote Sensing and GIS technology with important practical and theoretical value.This paper focuses on the updating of road databases, mainly involves with automatically or semi-automatic extraction strategy from images, and combined with this basis roads vector data to provide some automatic update routines. It mainly involves the following content:1) Based on the view of image segmentation, taking into account the road noise and the smooth distribution of non-road area, this paper introduces a method using level set image segmentation methods, including the use of partial differential equations for image smoothing and enhancement of regional roads, and the analysis of the characteristics of the level set method based on the combination of adaptive binary template matching method provides the initial value for the segmentation level set method, that is, the initial area of the road, then some post classify operation such as mathematical morphology operations or area threshold and last the result should be vectorised for output. 2) From the perspective of probability theory to express the features of road in image, this paper proposes an algorithm based on mean shift theory for extraction, in which the initial information is reduced to a single point and the width, the algorithm will automatically track the roads with adaptively changing the main direction of the roads when encountering large corners; and with the vector for the initial information an automatic extraction routine is also provided and compared with the classic ones,proves that it’s faster than the others.3) With the help of vector data, this paper proposed two work flows for the roads changes detection based on the point extraction from ortho images respectively, with using the combination of road extraction methods and the vector to provide important initial information, they will complete automatically. For some special situation, the extracted road points with more acute angles which are not agree with actual road distribution, so that a simple algorithm for smoothing these acute angles is provided.Ortho-image of 1:50,000 from Guangdong and Sichuan are the sources of this paper. They are used to prove the validity of the above methods for road extraction and the change detection of them. |
PDF Full Text Request