Bridge bolt detection method based on self-attention and central point regression model

Abstract

The invention discloses a bridge bolt detection method based on a self-attention and central point regression model. The method comprises the steps: acquiring a to-be-detected bridge bolt image; detecting the bridge bolt image through a pre-trained model based on self-attention and central point regression; and according to a detection result of the self-attention and central point regression model, determining whether the bridge bolt image bolt has a disease or not, wherein the model based on self-attention and central point regression comprises a semantic segmentation network and a detection network at the same time, the two networks share middle-layer features extracted by the convolutional neural network, the semantic segmentation network aims to perform semantic segmentation on a rectangular area where the bolt is located, and features for semantic segmentation and features of the detection module are connected to jointly position the bolt. The method is applied to the field of bolt disease detection and recognition, can overcome the defects of a traditional bridge disease image detection and recognition technology, and can well solve the problems of efficiency, cost, safety and the like in bolt disease detection and recognition.

Description

technical field [0001] The invention belongs to the field of vision technology, and relates to the application of vision technology in bridge diseases, in particular to a bridge bolt detection method based on self-attention and center point regression model. Background technique [0002] High-strength bolts are one of the main connection methods for large-scale steel structures such as bridges. The steel used for high-strength bolts in bridges in my country has developed from 40B to 20MnTiB and 35VB. Can meet the requirements of use. In recent years, affected by various factors, the probability of delayed fracture of high-strength bolts has increased. In operation, railway bridges have long been subjected to the design and operation load, speed-up, heavy-duty trains, natural disasters such as earthquakes, floods, mudslides, and harmful substances such as wind and rain, ice, and harmful ions. Accidental impact, etc., may cause different degrees of damage to piers and foundat...

AI Technical Summary

Problems solved by technology

[0006] (2) When the accuracy of auxiliary tools decreases or is damaged, it is inconvenient to replace and the cost is high;

[0007] (3) It cannot be automated and requires a lot of manual participation

[0011] (3) When constructing a key point heat map (heatmap), if you directly set the center point pixel to 1 and the remaining pixels to 0, it will be difficult to calculate the error due to the discontinuity of the optimization target, and then it will be difficult to optimize the entire image by error backpropagation. Model

Some people use the Gaussian kernel to disperse the center point on the entire heat map. Although some effects have been achieved, the foreground image of the generated heat map cannot effectively reflect the shape of the bolt because the calculated coordinate range is the entire heat map.

Images