To recognize large bridges over water under complicated background from optic images of natural scenes, a novel two step knowledge based strategy was presented. First, all regions in accord with the contextual description of the scenarjo about bridges over water, regions of interest, were detected from the field of view. Then, image segmentation and feature extraction were performed over all sub images of the detected regions of interest. Finally, based on the geometric characters of bridges, targets were identified by utilizing fuzzy reasoning approach and either the central pier of the bridge or the middle of the bridge body was selected as aimpoint furthermore. Experiments were conducted over different optic images of different kinds of large bridges over water in natural scenes, and the results demonstrated that the approach brought forth herein performed well in recognition accuracy and computational efficiency.