Corrosion detection method of substation Aboveground Steel Structure Based on deep learning

Aiming at the problems of time-consuming and low accuracy of the existing corrosion detection methods of substation above ground steel structures, a corrosion detection method of substation above ground steel structures based on deep learning is proposed. Firstly, preprocess the data set from the network and manual collection, expand the data set by translation transformation and cutting, and cut the original image; Secondly, aiming at the lack of fitting ability of shallow neural network model under small sample data, a double jump depth residual network model is established, and the double jump connection is introduced to enhance the corrosion feature detection ability of the model. Finally, the second-order moment estimation gradient affecting the learning rate in Adam algorithm is adjusted to speed up the corrosion detection speed of the proposed method. The experimental results show that the Precision, Recall, CAR and F1 scores of the proposed above ground steel structure corrosion detection method are 0.932, 0.923, 0.917 and 0.927 respectively, which are better than the compared methods. Therefore, the proposed method has good corrosion detection ability of above ground steel structures.