Quadcopter-Centric Crack Detection: YAML Algorithm Integration

Abstract

The primary objective of this project is to design and create a wall crack detecting system that uses the Yet Another Markup Language (YOLO) v81 model and an aerial drone. Wall fissures present serious maintenance and safety issues for power plants, buildings, and infrastructure. The YOLOv3 model, which is renowned for its real-time object detection abilities, is a major component of the project. By incorporating this model into an aerial drone platform, the system's mobility and accessibility are increased, making it possible for it to inspect areas of buildings and structures that are difficult to reach. The general perspective offered by the drone's aerial perspective makes it possible to identify fractures early on that could otherwise go undetected. The training process is guided by the YAML configuration, which includes important parameters for the best model performance, like epochs, batch size, picture size (imgsz), and pretraining. The computing core for onboard deployment is one example of how the system is designed for integration. Important factors to consider are labour assignments, device management, and project-specific information specified in the YAML file. The objective of this project is to address real-time processing and remote monitoring challenges in infrastructure inspection situations and to develop a scalable and reliable system for automated wall crack recognition.