Knowledge-Based Bridge Recognition in High Resolution Optical Imagery

Bridge detection in remote sensing imagery has been a topic of considerable interest. In this paper, a new flow of bridge recognition in high-resolution optical imagery is proposed. Based on the prior knowledge of bridge, potential river regions are segmented and thus the centerline of rivers can be extracted after morphology erosion and dilation operations. Then the bridge detector is implemented along the river centerline in ROI regions. After that, the detected bridges are identified by using two knowledge rules and the characteristics such as location, length, width and orientation of detected bridges can be obtained. Experiments are executed on the SPOT-5 5m-resolution panchromatic imagery, and the results are showing to confirm the validity of the proposed method.