Roads Extraction and Their Changes Detection in Airborne Images with the Help of Gis Data

This paper presents a novel routine to detect the changes of main roads with airborne images of medium-scale of 1:10,000. It mainly focuses on the changes detection of primary streets not the extraction of roads though it sounds an essential step. Here comes the brief work flows of this method: first, auto dodging the original image with Wallis Filter (Deren Li, 2006) as the pretreatment, second, get the variance and gradient images for the seeking of roads seeds, and third, a special searching processing is employed to locate the precise positions of the roads seeds in the above result and the original image in a special extension defined by the poor GIS data, and last, with several assumptions based on the gray properties of the roads in the image, preliminary changes of the primary roads can be detected. Results of all the steps mentioned above should have been refined by the GIS (Geographical Information System) data—the so-called DLG (Digital Linear Graphic) data, unfortunately, the DLG data at hand is too poor to identify the roads’ locations but to provide only initial positions of the main roads while the secondary access ones have to be ignored. Experiments prove that this work flow effective for roads with homogeneous gray distribution.