Super-Resolution Based and Topological Structure for Narrow Road Extraction from Remote Sensing Image

摘要

It is difficult to extract a narrow road with a width of only a few pixels from a remote sensing image. In order to solve this problem, it is proposed to process the remote sensing image with super-resolution. This paper extends the details of narrow roads by using the Deep Convolutional Neural Network (DCNN) method. Next, some noise points or roads of error extraction are processed by topological structure. To verify performance of the experimental method, experimental research on open remote sensing image data set is carried out. The experimental result is to compare the original image, super-resolution image, and topology filtering. Experimental results demonstrate that the new method has better effectiveness and superiority over the original remote sensing image.