Various traditional methods have been applied to detect cracks, these are prone to errors, require high human capital, high cost of equipment is involved and it is time consuming. In recent times, more sophisticated technologically based methods have been developed and used in the detection and analysis of cracks, inclusive of more advanced Artificial Intelligence (AI) technology. This study is aimed at evaluating the AI based Crack detection software for crack analysis. Cracks images were identified and captured using an android based smart phone from some selected residential buildings within Abuja and Benin-City. These crack images were analysed using Automated Building Exterior Crack Inspection Software (ABECIS). The software produced image segmentation and categorization. After the analysis, it consequently generated a report containing the confidence type, crack type, maximum confidence score, percentage of crack coverage and crack length. The cracks lengths for the selected images were also measured manually. The results indicate discrepancies between the crack lengths estimated by the software and those measured manually. The calculated Mean Absolute Error (MAE) is 0.84 while the Root Mean Square Error (RMSE) is 1.02. It was observed that greater error occurred in measuring the lengths of horizontal crack type. Despite these variations, ABECIS can significantly detect and measure the lengths of cracks to a reasonable degree of accuracy. The use of ABECIS saves time, cost, better analysis and portable record keeping for future review and maintainability of structural works for sustainable society. Therefore, it is recommended that the use of AI technology using ABECIS software can be advanced and promoted within the Nigerian construction industry for crack detection and analysis for effective maintainability and sustainability.
Key words: Crack detection, Crack analysis, Artificial intelligence, Image segmentation, Structural health monitoring.
|
