Crack identification method based on deep learning

Abstract

The invention provides a crack identification method based on deep learning, and belongs to the technical field of image processing, and the method specifically comprises the steps: constructing a deep convolutional adversarial network; acquiring a plurality of real crack image samples to train the deep convolutional adversarial network to obtain an adversarial crack image sample; obtaining a mixed crack image sample; labeling all the mixed crack image samples, and taking the labeling information file and the mixed crack image samples as a training set; training the improved YOLOv4 neural network by adopting the training set until the network converges, and storing convergent network parameters as a crack identification model; and inputting the acquired to-be-detected crack image into the crack identification model, and outputting identification information. Through the scheme of the invention, the deep convolutional adversarial network is constructed to realize the automatic generation of the crack, and the crack recognition model is obtained through the improved YOLOv4 neural network training to recognize the to-be-detected crack image, so that the recognition efficiency, accuracy and adaptability are improved.

Description

technical field [0001] The present application relates to the technical field of image processing, and in particular to a crack recognition method based on deep learning. Background technique [0002] At present, with the steady improvement of my country's economy, the speed and scale of infrastructure construction have reached a new climax. During the service of infrastructure, due to the joint influence of long-term loads, geological disasters, human activities and other internal and external factors, it is inevitable that various structural diseases will appear, among which cracks appear most frequently. If the treatment is carried out, it is very likely to cause hazards such as water leakage, concrete corrosion, and reduction of structural bearing capacity, which will affect the safety of the structure during operation. Therefore, it is necessary to regularly detect cracks in the infrastructure, grasp the number and specific locations of cracks, and take corresponding m...

AI Technical Summary

Problems solved by technology

[0006] In view of this, the present application provides a crack identification method based on deep learning, which at least partially solves the problems of poor identification efficiency, accuracy and adaptability in the prior art

Images