Knowledge-Based Identification and Damage Detection of Bridges Spanning Water via High-Spatial-Resolution Optical Remotely Sensed Imagery

Abstract

Bridges over water are important artificial objects that can be damaged by natural disasters. Accurate identification and damage detection of such bridges through the use of high-spatial-resolution optical remotely sensed imagery are important in emergency rescue and lifeline safety assessment. In this study, we detail a knowledge-based method of identification and damage detection of bridges spanning water using high-spatial-resolution optical remotely sensed imagery. Data on the body of water are extracted to define spatial extent and improve the timeliness of identification and damage detection, the threshold values of the rectangle degree and area are set to remove false bridge targets, and the damaged parts are detected according to the bridge’s rectangular characteristics and the relationship with the body of water. First, the characteristics, such as spectral, geometric, and textural, and spatial relationships of the bridge over water, are analyzed. Second, to limit the spatial extent of bridge identification and improve computational efficiency, data on the body of water are extracted. Third, the post-event bridge is identified from the viewpoint of bridge integrity based on shape and area parameters. Damage detection is then performed according to the bridge’s integrity. Finally, the results are evaluated for both non-positional and positional accuracy. Results of experiments carried out in Huiyang and Wenchuan, China, show that the proposed method, using high-spatial-resolution optical remotely sensed imagery, is effective for identification and damage detection of fallen and collapsed bridges spanning water. Therefore, the method is useful in updating the geographic database of bridges and assessing damage to them caused by natural disasters.