Automated Detection of Structural Anomalies Using Object Tracking Techniques

Computer Engineering

Buildings, pavements, and bridges, among others - are the cornerstones of modern infrastructure. These are exposed to degradation from various environmental conditions, material fatigue, and usage patterns. This often evolves into developing defects over time. Therefore, early detection of these anomalies is critical in ensuring safety, longevity, and economic viability. This is a very important task of civil engineering as detection of defects in time may prevent probable failures. Common anomalies include cracks in buildings, potholes in pavements, corrosion in bridges, and other structural defects. Inspections by the conventional approach are manual, time-consuming, and liable to human mistakes. An automated and efficient approach must be identified to detect these anomalies at an early stage. This project develops an automated system that detects anomalies in buildings, pavements, and bridges in civil engineering structures early, using advanced image segmentation techniques. We first collect images from the web and then create a custom dataset to train and evaluate segmentation models like YOLO and Detectron2. Such models will be fine-tuned for the more accurate identification and segmentation of structural anomalies so that maintenance interventions can be implemented on time and with accuracy. The implementation of image segmentation will not only increase the precision and efficiency of defect detection but will also improve the safety and durability of civil engineering structures. This project aims to provide infrastructure management with an extentable solution, given the automation of the inspection process in order to detect potential issues at an early stage and also extend the lifetime of such structures.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.