Autonomous UAV Navigation and Real-Time Crack Detection for Infrastructure Inspection in GPS-Denied Environments

Manual inspection remains the primary method for detecting structural defects in domestic infrastructure. However, it often struggles to meet the requirements of modern infrastructure inspection in terms of accuracy and speed due to the complexity of the systems and the subjective judgment of inspection personnel. In addition, the intricacy of infrastructure images presents significant challenges to traditional image processing methods. Uncrewed aerial vehicles (UAVs) have proven to be effective in identifying structural damage across various types of infrastructure. Nevertheless, infrastructure inspections present unique challenges, including the absence of GPS signals and high collision risks in specific environments. This article introduces a GPS-free self-localization algorithm that utilizes onboard sensors to move autonomously and a segmentation neural network for precise defect detection to address these challenges. Field tests have confirmed the framework's effectiveness, highlighting its potential for broader application in infrastructure inspections. This innovative approach has the potential to enhance inspection accuracy, efficiency, and safety, thereby ushering in more reliable and automated infrastructure maintenance practices.