Abstract
Cracks classification on buildings caused by natural disasters such as earthquakes can be done manually by analyzing walls, beams, columns, and floors based on visual inspection of cracks diameter, depth, and length. The manual assessment method requires experts in structural engineering who have enough knowledge and experience in building damage assessment. To facilitate and overcome these problems, a crack classification system is developed by using a digital image processing approach (pattern recognition) that can classify cracks into the mild, moderate, or severe categories using GLCM features and SVM classifier. Based on the experimental results that the proposed method has appropriately worked for classification of two crack classes (mild and severe) indicated by 94.44% of accuracy, 88.89% of precision, and 100.00% of recall. While for three crack classes (mild, moderate, and severe) obtained the accuracy 81,48%, recall 81,41% and precision 88,09%. Furthermore, the proposed system also shows robust performance against large variability of crack and non-crack images, and the SVM classifier outperforms over the statistical-based classifier (LDA and QDA).
Concepts :
Access to Document
10.1109/icamimia47173.2019.9223403SDGs
Citations by Year
| Year | Count |
|---|---|
| 2019 | 5 |