Road segmentation on remote sensing images: aerial (or very high resolution) images and satellite (or high resolution) images, has been employed to various application domains, particularly road extraction in which the segmented objects are served as a mandatory layer in geospatial databases. Several attempts in applying deep convolutional neural network (DCNN) to extract roads from remotely-sensed images have been made; nevertheless, the accuracy is still restricted. This thesis presents an enhanced DCNN framework specifically tailored for road extraction on remotely-sensed images by applying landscape metrics (LMs) and conditional random fields (CRFs). To improve DCNN, a modern activation function; called exponential linear unit (ELU), is engaged in our architecture resulting in a higher number of and yet more accurate extracted roads. To further alleviate falsely classified road objects, a solution based on an adoption of LMs is proposed. Lastly, to sharpen the extracted roads, a CRF method is added to our framework. The experiments were conducted on Massachusetts road aerial imagery as well as Thaichote/THEOS satellite imagery data sets. The results demonstrated that our proposed framework outperformed SegNet, the state-of-the-art object segmentation technique on any kinds of remotely-sensed imagery, in most of the cases in terms of precision, recall, and F1 scores
