Authors: Miss Anukula Roja, Athili Venkat Raju
Abstract: Surface defect detection in strip steel is a critical quality assurance task in modern steel manufacturing, as defects such as scratches, inclusions, patches, and rolled-in scales can significantly degrade product quality, mechanical performance, and production efficiency. Conventional manual inspection methods are labour-intensive, subjective, and incapable of satisfying the speed and accuracy requirements of automated manufacturing environments. This paper presents an intelligent hybrid framework for automated strip steel surface defect detection by integrating traditional machine learning and deep learning techniques. The proposed approach incorporates mean filtering for noise reduction and adaptive threshold-based segmentation to accurately extract defect regions from strip steel images. To improve classification performance, an ensemble model combining Random Forest (RF) and ResNet50 is developed, where ResNet50 extracts rich hierarchical visual features and Random Forest effectively classifies discriminative statistical features. The proposed framework is evaluated using a multi-class strip steel surface defect dataset comprising various defect categories. Experimental results demonstrate that the hybrid RF–ResNet50 model outperforms individual machine learning and deep learning models in terms of classification accuracy, robustness, and generalization capability. The complementary learning characteristics of both models enable effective representation of low-level texture information and high-level semantic features, resulting in reliable defect identification under diverse surface conditions. Furthermore, the proposed framework is computationally efficient and scalable for real-time industrial deployment, reducing reliance on manual inspection while enhancing product quality and manufacturing productivity. These findings highlight the potential of hybrid artificial intelligence techniques as an effective solution for next-generation intelligent surface inspection and quality control systems in smart manufacturing.
DOI: http://doi.org/http://doi.org/10.61137/ijsret.vol.12.issue4.146
