Comparative analysis of deep learning and machine learning algorithm for concrete crack detection

• Main Authors: Padmavathy, R., Kalaiarasi, G., Venkatasubramanian, R., Devi, M., Grace, L. K. Joshila
• Format: Conference Proceeding
• Language: English
• Subjects: Algorithms; Artificial neural networks; Damage detection; Deep learning; Flaw detection; Machine learning; Support vector machines
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0222363

This research is focussed on detection of cracks in concrete surfaces. Existence of cracks shows that the surface is degrading and cause severe damage when it is continuously exposed to all climatic conditions. Generally an expert examines the crack but it is not surely known about the depth of the crack to be repaired. This research work can be the replacement and finds the spreading level of the concrete cracks accurately[2]. Image preprocessing techniques are used to extract the significant features of the input images. Classification algorithms such as Support Vector Machine and Convolutional Neural Network (CNN) are experimenting with the cracks features into different stages as initial level, medium level and advanced level[14]. CNN gives better accuracy, reduced training and testing time. By detecting the level-of crack, the necessary steps can be taken before any accident happens.