Crack Detection System Using Drone-Captured Images

Abstract

Structural cracks in concrete infrastructures such as buildings, bridges, and pavements pose significant safety and durability risks if not detected at an early stage. Conventional crack inspection methods rely heavily on manual visual assessment, which is time-consuming, subjective, and unsafe for large-scale or hard-to-reach structures. To address these limitations, this project presents a Drone-Based Concrete Crack Detection System using Machine Learning, integrated with an interactive Streamlit web application. The system utilizes drone-captured images as input and employs a Convolutional Neural Network (CNN) to automatically detect the presence of cracks. In addition, image processing techniques such as grayscale conversion, thresholding, and pixel analysis are applied to estimate crack severity quantitatively. The developed web-based interface allows users to upload multiple images, visualize detected cracks through overlay highlighting, and receive automated decision outputs. The system also generates downloadable CSV and PDF reports and provides audio-based result announcements to enhance usability. The proposed solution offers an efficient, scalable, and non-invasive approach for automated structural health monitoring and early crack detection.