<![CDATA[The growing demand for an accurate and rapid method to assess concrete strength has driven the development of non-destructive and cost-effective techniques. This paper aims to enhance the process of extracting shape, texture, and color features from concrete surface images to improve the accuracy of strength classification through digital image processing and artificial intelligence (AI). The study uses a dataset of 300 high-resolution photographs of concrete samples, categorized by their compressive strength levels: weak, moderate, and strong. These images were taken under controlled background and lighting conditions to ensure consistency. The methodology involves three stages: image preprocessing, feature extraction, and classification. During preprocessing, RGB images are converted to the Lab color space, and a three-layer median filter is applied to reduce noise. The K-Means clustering algorithm segments the images, and relevant features such as Metric, Eccentricity, Contrast, Correlation, Energy, Homogeneity, Hue, and Saturation are extracted. Among these, Correlation and Energy are the most influential in classification accuracy. The experimental results show that the proposed approach can reach up to 90 percent accuracy in classifying concrete strength into three categories. This suggests that visual features have strong potential to replace traditional destructive testing methods. The findings also point to the possibility of enhancing prediction accuracy with deep learning models and developing real-time, field-based evaluation tools to aid quality control in the construction industry.]]>
Copyrights © 2025
