Exploiting context for assisted aerial image interpretation
Alexander Bauer, Oliver Herschelmann · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2010
The evaluation of a country's critical infrastructure requires a detailed analysis of facilities such as airfields, harbors and heavy industry. To improve the assessment of such facilities, an assistance system for the interpretation of infrastructure facilities from aerial imagery is developed. In this paper we point out recent advances of the system's recommendation function. Besides suggesting the occurrence of undetected objects based on a probabilistic scene model and previously detected objects, the system is now able to suggest the classification of objects based on intrinsic object features and both local context (spatial relations) and global context (overall scene classification). To justify our approach the results of an experimental evaluation of the system for the classification of industrial installations is presented.