Use of Deep Learning Techniques for Road Extraction using Remote Sensing Imagery

Road Extraction plays a crucial role in a myriad of domains such as city management, traffic system management, and Global Positioning System (GPS). The amount of data that can be used for this purpose is significantly increasing with time because of improvements in data extraction and storage technologies. One of the most important sources for such data are the remote sensing images like from satellites and UAVs. However, these images can contain certain flaws such as blurriness that can make it difficult to extract the roads efficiently. To solve this issue, Deep Learning techniques have been proposed and utilized in various research publications. This paper presents the analysis and review of the machine learning techniques implemented in some of the publications in the recent years for the purpose of road extraction using remote sensing imagery. To overcome some of the common weaknesses, it then proposes the use of a deep learning model as one of the better solutions that can provide efficient results.