Abstract: In recent years, the bridge has been hit by collision accidents frequently, and emergency rescue needs to respond quickly. Detecting the damage form, damage location and damage degree of the bridge structure is a necessary link to evaluate the risk of secondary instability and collapse, and to provide a basis for the formulation of repair plans. The underwater environment of the accident site is complex, the risk of manual operation is high, the error of damage location is large, and the field of view of the water turbidity operators is limited, and the detection cannot be traversed, and the detection results are greatly affected by the experience of the inspectors. In this paper, two non-contact detection methods based on underwater robot video equipment and unmanned ship carrying multi-beam are proposed and verified. The damage of the bridge is detected from the detailed damage determination and the overall structural damage assessment. In this paper, the underwater robot can accurately locate the damaged position and effectively evaluate the damage degree. The three-dimensional point cloud model of pier column obtained by multi-beam can accurately measure the slope of pier column and judge the bending position of pier column. After detection, it is determined that the accident pier is inclined as a whole, and there is local damage but no bending damage. The mutual verification with the traditional displacement measurement data shows that the above two bridge non-contact detection methods are accurate and feasible. In the emergency rescue detection of bridge accidents, it has the advantages of timely response, avoiding personal injury, accurate positioning,and provide effective solutions for similar projects.