Road Network Extraction from High Temporal and Spatial Resolution Remote Sensing Data
At present, most of the road network extraction is based on the outdated machine learning algorithm, which cannot overcome the problem like instantaneous updating of the road network. Information about road networks not only important for transportation but also to respond to natural calamities emergency. The enormous evolution in observation of remote sensing data enabled in the mid-20th century and became a hot topic. This work proposed a method for automatic extraction of road along with path travel time and the speed limit for each street from high temporal remote sensing images. By combining Atrous Spatial Pyramid Pooling(ASPP) with an encoder-decoder network, the efficiency of the road extraction network also enhanced. The proposed method allows use of ASPP's multiscale feature extraction capabilities as well as the Encoder-Decoder network's comprehensive feature extraction capabilities. As a result, it can provide precise and informative road extraction results. Optimal routing is also feasible with this method. The result indicating that the proposed method can be used in complex scenes and extract road networks clearly and efficiently.
Keywords-Natural calamities, remote sensing, Encoder-Decoder, optimal routing.