Abstract: A method for thinning and connection in linear object extraction from an image and including the following steps: 1. extracting direction features using various sliding windows from the binary image obtained of linear objects. 2. decomposing the binary image into several binary image layers according to the direction features. 3. extracting thinned curves and endpoints of each binary image layer by conducting curve fitting on each connected component using coordinate information. 4. connecting the thinned curves by computing spatial distances between the endpoints belonging to different thinned curves, angles between the tangential direction and connected direction vectors of the connected points. Finally, the road network image is constructed by overlaying image layers with the thinned curves.

Abstract: A method for thinning and connection in linear object extraction from an image and including the following steps: 1. extracting direction features using various sliding windows from the binary image obtained of linear objects. 2. decomposing the binary image into several binary image layers according to the direction features. 3. extracting thinned curves and endpoints of each binary image layer by conducting curve fitting on each connected component using coordinate information. 4. connecting the thinned curves by computing spatial distances between the endpoints belonging to different thinned curves, angles between the tangential direction and connected direction vectors of the connected points. Finally, the road network image is constructed by overlaying image layers with the thinned curves.