Accurate Detection Method of Corrosion State of Steel Structure Bridge in Internet of Things Environment Based on Visual Image Characteristics

Abstract

Abstract Bridge structures are one of the most important aspects of transportation because they make remote areas accessible, but preserving the environment is equally important. The toughness and endurance of the bridge structure is very important from the security perspective of transportation. Corrosion adversely impacts the steel structure strength of bridges. Accurate detection methods within the environment of Internet of Things can help to find the corrosion of bridges in time, take maintenance measures in advance, and delay the decay of bridge life. At present, the inspection of bridge supports is primarily carried out by labor-intensive inspection. This method is time-consuming and labor-intensive and also affects traffic. To show advancement in the detection accuracy of the bridge corrosion state, an accurate detection method based on visual image features is proposed. Drone technology is used to collect corrosion images of steel bridges. Considering the complexity of the image, the convolution operation is performed on the images using a deep neural network (DNN). A DNN model is constructed according to the apparent features of the rust image. The supervised learning DNN is combined with the unsupervised learning sparse autoencoding (SAE), and the DNN is autoencoded by SAE to reduce the reconstruction bias. On this basis, the accurate detection of the rusted state is accomplished. From the experimental analysis, it is apparent that the peak signal-to-noise ratio of this method is higher than 25, and the detection time is shorter than that of the methods compared. It can accurately detect different types of rust states.