Intelligent Detection Technology for Concrete Cracks in Bridge Baseplate Based on Machine Vision

Abstract

With the continuous advancement of urbanization, the demand for the use of bridges has increased, and the demand for bridge health monitoring also sharply increased. Because of the particularity of its location, there are many difficulties in the crack detection of bridge baseplate. This paper proposed an intelligent detection technology for concrete cracks in bridge baseplate based on machine vision. Firstly, the images of concrete crack in bridge baseplate were collected by unmanned aerial vehicle (UAV), the images were binarized, and the binarized images were used as the data set. Then, we improved the VGG model, proposed the VGG-A9 model suitable for crack image classification, set the ReLU function, added the batch normalization layer and dropout layer to optimize the model. The VGG-A9 model was trained and verified through the constructed data set, the loss function curve and classification accuracy curve were analyzed. It can be seen that the VGG-A9 model has low loss value, high accuracy and fast convergence speed. Finally, the crack feature parameters were extracted from the classified crack binary images as the basis for the service reliability evaluation of bridges, which provide reference for the intelligent detection and operation maintenance of bridges.