Bridge surface disease automatic identification system

Abstract

The invention discloses a bridge surface disease automatic identification system, which acquires a bridge surface image, marks a bridge part to which the bridge surface image belongs as a label L1, and marks a bridge surface disease category and image coordinate information as a label L2 according to the bridge part; randomly divides the label L1 into a training set and a verification set in proportion, and constructs a classification network model; randomly divides the label L2 into a training set and a verification set in proportion according to bridge parts, and constructs a target detection network model; inputs the bridge surface image I into the trained classification network model to obtain the confidence coefficient belonging to each type of bridge parts, and takes the bridge part corresponding to the maximum confidence coefficient as the bridge part where the bridge surface image I is located; and inputs the bridge surface image I into the target detection network model of the bridge part, and outputs the predicted disease category, category confidence and image coordinate information. The technical problems that an existing bridge disease recognition device supports a single disease type and is not high in accuracy can be solved.

Description

technical field [0001] The invention relates to the technical field of rail engineering machinery, in particular to an automatic identification system applied to bridge surface disease detection. Background technique [0002] During the service of the bridge, it is inevitable that there will be cracks, exposed tendons, honeycomb pitting and other diseases on the surface. The existence of these diseases makes the bridge a potential safety hazard. Regular inspection of the bridge surface in order to detect the disease in time and judge the risk level of the disease is an effective means of monitoring the health of the bridge, which is directly related to the operation safety of railways and highways. The traditional bridge disease detection method is that inspectors use telescopes to observe bridge bottom diseases at a long distance, or observe bridge bottom diseases at close range through the built bridge bottom observation platform. The prior art uses inspectors to manuall...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the purpose of the present invention is to provide a bridge surface disease automatic identification system to solve the technical problem that the existing bridge disease identification device supports single disease types and low accuracy

Images