Research On Road Network Change Detection Method Based On Remote Sensing Image

As one of the important objects in remote sensing image,road plays a very important role in the research of national economy and geographic information.The road network is composed of various road objectives,interwoven into a network of road system.For the constantly changing and updating road information,automatic and efficient remote sensing image road network change detection technology has always been the research hotspot and difficulty in the field of photogrammetry and remote sensing.The research of road network change detection method is of great significance to road monitoring,map updating and weather disaster detection.Based on this background,thesis carried out the research on road network change detection technology based on remote sensing images.The main contents are summarized as follows:(1)Remote sensing image and road network preprocessing process for road network change detection are established.Remote sensing image preprocessing includes image target area clipping and coordinate system conversion.Road network preprocessing includes road network refinement and road network vectorization.A road network refinement method combining Zhang-Suen(ZS)refinement algorithm and table lookup method is proposed in thesis.Experimental results show that this method can significantly improve the quality of refining results.In order to reduce the amount of road network data,thesis further carries out vectorization operation on the refined raster road network data,and the generated vector road network will be used as the data input for subsequent change detection.(2)To reduce the influence of image registration on road network change detection performance,an improved Scale-invariant feature transform(SIFT)algorithm is proposed to realize efficient registration of remote sensing image.Aiming at the problems of the traditional SIFT algorithm in the registration process of remote sensing image,such as the uniqueness of feature descriptors,low robustness,and the number of correctly matched points with the same name,thesis combined Sobel edge detector,variable gradient position direction histogram and improved random sampling consistency algorithm to optimize the traditional SIFT algorithm.Experimental results show that the proposed optimization method can significantly improve the uniqueness and robustness of feature descriptors,and increase the number of correctly matched points with the same name.(3)Aiming at the problem of vector road network change detection,combined with buffer growth method technology,a road network change detection algorithm based on improved Fraser distance model in remote sensing image was proposed to detect road network change information from the whole to the part.In order to avoid mismatching caused by matching confusion,the method assigns directional attributes to all local road segments,and divides the change detection results into three parts: unchanged road,dismantled road and newly added road according to the source of the changed road.(4)Design and develop the prototype system of road network change detection in remote sensing images.The system includes image basic operation module,remote sensing image registration module and road network change detection module,which realizes road network change detection operation conveniently and quickly.