Object-Oriented Extraction of Highway Bridges Using Very High Resolution Digital Aerial Images and Digital Surface Model

Remotely sensed images play an important role in spatial data production and defence applications. Turkey has a significant number of ground observation platforms, the number of which is increasing every year. Consequently, the number of images that need to be processed has increased significantly. Processing these images manually causes loss of time, money, and labor. Therefore, it is necessary to process these images using automatic/semi-automatic methods. Information can be extracted from images via image classification or feature extraction. Thus, different traditional/nontraditional methods have been used to extract information from various images. However, it is not possible to extract objects using only the images. An example of such an object is a highway bridge that is not over water. In terms of Geographic Information Systems (GIS), a meaningful object and as they are a critical component of transportation networks-highway bridges-it is necessary to extract highway bridges. Object-oriented methods have been employed to extract highway bridges using digital aerial images and digital surface models (DSM). Firstly 12 cm resolution digital aerial images were used to produce the DSM. The study area was classified into vegetation and non-vegetation areas. To extract highway bridges, fuzzy rule sets have been developed using eCognition software. An accuracy assessment was then performed. All highway bridges (six) in the study area were successfully extracted.