Font Size: a A A

Research On Approaches Of Recognizing Urban Road From High Resolution Satellite Optical Imagery

Posted on:2007-01-14Degree:DoctorType:Dissertation
Country:ChinaCandidate:H LongFull Text:PDF
GTID:1118360185478897Subject:Cartography and Geographic Information System
Abstract/Summary:PDF Full Text Request
Aimed at the high resolution remote sensing imagery, this thesis firstly analysized the main properties of the different rank roads in the imagery, and classify the roads into two kings: main-width roads and the secondary-width roads according to the pixels' width of the roads in the imagery and based on the ideal road model, secondly, present approaches to detect the roads from the imagery, and developed the according key algorithms which can detect the main roads and the secondary roads precisly.The contents of this thesis mainly consist of three parts: 1) improvement the image at the pre-process stage, which include the image segmentation and the exclusion of the small obstacles in the imagery; 2) detection of the road-seeds for the main-width roads and tracking the roads by image matching techniques to form the road networks; 3) detection of the secondary -width roads to form the networks.At the pre-process stage, to improve the image, the method of gradient of the density in the feature space is used to detect the maximum-density position (the mode position) within a local space. The imagery is segmented based on the K-mean classification of the mode points which index the according pixels in the imagery. During these processes, a new adaptive approach to select the shape and the sizes of the Parzen window is presented.The main-width roads are detected based on the road ideal model and the result of the segmentation of the imagery. Five typical region geometric factors are selected to detect the typical road segments which are used to create road-seed pairs from the segmented imagery automatically. By using the orientation, the position, and the grey properties of the roads, we developed a road-tracing algorithm which matches the small patches' wavelet moments around the road seed. And these approaches detected the main road networks effectively.The detection of the secondary-width roads consist of detection of the center points (lines) and formations of the networks from the original center line-likes. In the process of detecting the center lines of the secondary-width roads, the secondary-width roads are considered as the continuous curve structures like ridges in...
Keywords/Search Tags:Road extraction, Gradient of feature density, Road tracing based on image matching, Second derivative ridge, Stochastic point process
PDF Full Text Request
Related items