Intelligent Extraction of Surface Cracks on LNG Outer Tanks Based on Close-Range Image Point Clouds and Infrared Imagery

The article introduces an innovative method for detecting surface cracks on LNG outer tanks using infrared imagery and 3D point cloud data. The approach combines convolutional neural networks with self-attention mechanisms to enhance crack detection accuracy. The method involves capturing infrared images and close-range images using UAVs, aligning them with 3D point cloud data, and applying a convolutional neural network to identify cracks. The research highlights the integration of thermal infrared images and point clouds to mitigate environmental interference, providing a robust solution for detecting cracks that are not visible to the naked eye. The study demonstrates the effectiveness of this method through experimental results, showing high accuracy in crack detection and extraction. The article concludes by emphasizing the importance of this technology for ensuring the safety and reliability of LNG tank structures.