Automated Crack Detection And Quantification In Concrete Surfaces Using Mask R-Cnn

Abstract

Crack detection in concrete structures is a critical aspect of structural health monitoring and maintenance management. Traditional inspection methods are labor-intensive, subjective, and time-consuming. Recent advances in deep learning have enabled automated crack detection with improved accuracy and efficiency. This research presents an automated crack detection and quantification framework based on Mask Region Based Convolutional Neural Network (Mask R-CNN) for identifying and measuring cracks in concrete surfaces. The proposed model performs pixel-level segmentation of cracks, enabling accurate quantification of crack length, width, and affected area. Experimental evaluation on a dataset of concrete surface images demonstrates superior performance with a detection accuracy of 96.8%, precision of 95.7%, recall of 94.9%, and mean Intersection over Union (mIoU) of 91.6%. The results indicate that Mask R-CNN is an effective tool for automated structural inspection and maintenance planning.