A fuzzy-logic classifier for estimating the reliability of the self calibration of an embedded stereovision system
Sébastien Kramm, Pierre Miché, Abdelaziz Bensrhair · 2008
Estimation of epipolar geometry can be done automatically in on-board stereovision systems, using interest points that are detected and matched. However, image disturbance that can happen in real-life situations can considerably lower the performance. A reliability score computing method is proposed, based on a fuzzy logic classifier. Its input is the data extracted from the estimation process. The classifier is trained with artificial image disturbance, using a set of typical image pairs. Results show that the computed score is indeed related to the performance of estimation.