A Deep Learning-Based Approach for Pipeline Cracks Monitoring

In order to improve the efficiency and accuracy of pipeline surface cracks monitoring based on image processing, the Convolutional Neural Network (CNN) algorithm in target detection is introduced to quickly identify the type, location, and area for the extracted cracks area with borders, the CNN based on crack contour network (CCN) method used to locate and extract the crack shape. CCN algorithm introduces the accuracy rate (P%), recall rate (R%), and F-score (F%) index to evaluate the algorithm in the problem during cracks monitoring, and determines the corresponding contour area of the crack frame according to the maximum F-score. A pipeline image was carried out by using an inspection drone with high definition camera. The results show the recognition efficiency and accuracy of the proposed method. After the optimal value of the degree threshold, the accuracy rate, recall rate, and F-score are recorded 91. 8%, 86. 1%, and 84.6%, respectively.