Improving Planimetric Mapping Accuracy for Urban Areas from Aerial Imagery Using Geometric Constraints

Traditional terrestrial surveying techniques in mapping urban areas are usually inadequate for many reasons. For example, road networks always suffer large volumes of vehicles and personnel mobilities that affect road map production. Accordingly, both airborne and spaceborne techniques help so much in this issue, although defining and segmenting road regions and edges from these remotely sensed data will be kept as a challenging task due to large variations on road surfaces. The corresponding developments in these techniques include imagery systems, imagery platforms and imagery processing. The paper addresses this motivation by increasing the horizontal accuracy of output roads map from airborne imagery system. In this context, a stereopair of aerial images (0.80 m GSD) are available for a certain study area within many road segments and edge varying in lengths and heights. This stereopair is initially processed through the well-known conventional bundle adjustment algorithm using the commercial ERDAS Imagine 2016 software. Then, some geometric constraints concerning the straightness, curvature and intersections of tested edges are implemented and added in a new developed MATLAB program. The output results show the verification of the developed program when compared with the commercial one, besides the used constrained bundle adjustment yields to better accuracy reaches up to 45%.in the final produced road map.