Full-process bridge detection method based on deep learning

The invention relates to the field of computer vision and image processing, in particular to a full-process bridge detection method based on deep learning, and the method specifically comprises the following steps: S1, recognizing the type of a bridge from a long-distance image through employing an image classification algorithm, and carrying out the classification; s2, according to the identified type, automatically selecting and loading a target detection model corresponding to the large-volume component of the bridge, and realizing accurate identification of a middle distance and positioning of the large-volume component; s3, carrying out disease segmentation and identification, and providing a pixel-level disease contour; s4, dynamically adjusting the scoring weight, and providing a structured weight adjustment scheme for disease scoring; and S5, calculating a risk score by using a thermodynamic statistical algorithm according to the importance of the component corresponding to the disease features, and v