Road Segmentation

Semantic segmentation of remotely-sensed aerial (or very-high resolution, VHS) images and satellite (or high-resolution, HR) images has numerous application domains, particularly in road extraction, where the segmented objects serve as essential layers in geospatial databases. Despite several efforts to use deep convolutional neural networks (DCNNs) for road extraction from remote sensing images, accuracy remains a challenge. This paper introduces an enhanced DCNN framework specifically designed for road extraction from remote sensing images by incorporating landscape metrics (LMs) and conditional random fields (CRFs). Our framework employs the exponential linear unit (ELU) activation function to improve the DCNN, leading to a higher quantity and more accurate road extraction. Additionally, to minimize false classifications of road objects, we propose a solution based on the integration of LMs. To further refine the extracted roads, a CRF method is incorporated into our framework. Experiments conducted on Massachusetts road aerial imagery and Thailand Earth Observation System (THEOS) satellite imagery datasets demonstrated that our proposed framework outperforms SegNet, a state-of-the-art object segmentation technique, in most cases regarding precision, recall, and F1 score across various types of remote sensing imagery.

In this paper, we introduce an improved deep convolutional encoder-decoder network (DCED) for segmenting road objects from aerial images. Enhancements include the use of ELU (exponential linear unit) instead of ReLU, dataset augmentation with incrementally-rotated images to increase training data by eight times, and the use of landscape metrics to remove false road objects. Tested on the Massachusetts Roads dataset, our method outperformed the SegNet benchmark and other baselines in precision, recall, and F1 scores.

Road map extraction is vital for GIS and underpins many location-based applications like GPS navigation, delivery route planning, tourist attraction locating, and location-based marketing. This research uses satellite imagery, though other remotely sensed images like aerial photographs, UAVs, or drones are also applicable. Despite various proposed methods focusing primarily on accuracy, completeness of results is equally important. We enhance accuracy by incorporating connected component analysis and improve completeness using landscape metrics, which describe spatial characteristics through shape and isolation indices. Evaluated on precision, recall, quality, and F1 scores, our method achieves over 90% performance in all criteria.

Data extraction of geo-spatial objects from satellite images is a crucial step in facilitating analysis of geo-spatial or spatio-temporal data, typically involving line (road) and polygon (area) layers. This paper introduces a method for transforming satellite data (raster images) containing roads from pixel form into spatial objects comprising lines and polygons. Our algorithm consists of three primary steps. First, roads are isolated from other objects using k-means clustering. Second, line extraction is performed on the road areas by applying morphological operations to skeletonize the image, followed by enhancement using the Ramer-Douglas-Peucker algorithm. Finally, land-cover classification is applied to non-road objects to extract polygons. Experimental results demonstrate that both lines (road networks) and polygons (areas) can be accurately extracted from satellite imagery simultaneously.