Research on Concrete Crack Recognition and Classification Algorithm Based on Neural Network

This study proposes a method for identifying and classifying concrete cracks based on convolutional neural network (CNN), Inception module and Transfer Learning. A novel Convolutional Inception Transfer Network (CITN) model is constructed. Firstly, the local features of concrete images are extracted by CNN to capture the details of cracks. Secondly, the Inception module utilized multi-scale convolution operations to enhance the model's ability to extract crack features at different scales. Finally, by incorporating transfer learning techniques, the network parameters were initialized with a pre-trained model, which improved the model's generalization capability and training efficiency. In the experimental part, we compared with traditional methods, and the results show that the classification accuracy of CITN model on multiple concrete crack data sets is significantly higher than that of traditional methods. Especially in the case of limited data, CITN model has more obvious advantages, providing an efficient and accurate solution for the identification and classification of concrete cracks. This result fully validates the reliability and effectiveness of CITN architecture in the identification and classification of concrete cracks.