Automated Crack Detection and Severity Analysis Using Image Processing

Abstract

Monitoring of structural health is imperative for the proper working and life span of structures. One of the most significant signs of deterioration of such structures is cracks formation in the structures. Manually inspecting such cracks often consumes much time and lacks objectivity, resulting in poor inspection efficiency. In this paper, we propose an automatic method for crack detection based on the use of image processing techniques with Convolutional Neural Network (CNN). Image acquisition, followed by preprocessing through grayscale transformation and noise removal, and segmenting the image are done to detect cracks from the processed image with a higher degree of visibility. Finally, the processed image is classified into two categories based on whether it contains cracks or not using a trained CNN.