Speaker
Description
Timely detection of cracks in structural elements is essential for ensuring the safety and durability of civil infrastructure. Traditional inspection tech-niques, such as manual visual assessments, are often labor-intensive, subjec-tive, and ineffective for large-scale or hard-to-access structures. This paper presents an automated crack detection framework based on Haralick texture features extracted from Grey-Level Co-occurrence Matrices (GLCM). The proposed methodology includes image preprocessing, feature extraction, fea-ture selection, and classification. Evaluation is conducted using the publicly available SDNET 2018 dataset, which contains concrete surface images cap-tured under diverse lighting and crack conditions. To enhance computational efficiency, redundant features are eliminated using Ridge and LASSO regres-sion techniques. Support Vector Machines (SVM) with various kernel func-tions are employed for classification, validated through a 5-fold cross-validation strategy. Experimental results indicate that the proposed method achieves over 95% accuracy using a subset of selected features, demonstrat-ing its effectiveness and robustness for crack detection in concrete struc-tures.