A new method for observer-based evaluation of object detectability using image sequences taken by approaching sensors
E. Mauer, P. Koffler · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2008
Detectability evaluation of objects is required for sensor range performance determination or for the assessment of camouflage effectiveness. In order to obtain a statistically sound statement the evaluation is performed in a sequential or parallel manner by different human observers analysing stored image sequences, so called "video simulation trials". These sequences are captured by high quality sensors approaching the objects of interest. The proposed method is based on the visualization in real-time of different kinds of a priori knowledge related to the object location in every image of the image sequence in order to reduce the searching process of the human observer depending on recognition task. Hence the geometric transformation of geometric information through every image of the sequence is performed by applying an automatic image registration method, planar homography. Additionally, knowledge of these geometric transformations allows new trial options and trial automation so that further experiments for the proposal of a new evaluation standard for video simulation can be performed with a statistically significant number of observers.