An Image Recognition Algorithm for Defect Detection of Underground Pipelines Based on Convolutional Neural Network

The proliferation of underground pipelines is a defining feature of urbanization. Regular inspection and maintenance are necessary to reduce the economic loss caused by pipeline defects. This paper aims to detect pipeline defects in video images with the aid of computer vision. Firstly, the recursive neural network (RNN) was added to the classic convolutional neural network (CNN) to acquire various features from the images. Then, the Fisher criterion was weighted and improved, and introduced to the least square error cost function, enhancing the recognition rate of the improved CNN. Finally, the improved CNN algorithm was verified through contrastive experiments on actual underground pipeline images. The research results shed new light on the defect detection and maintenance of underground pipelines in urban areas.