DETECTION OF MATERIAL SURFACE DEFECTS USING DISTRIBUTED IMAGE INTENSITY FEATURES

Abstract

A method has been proposed and investigated that transforms images of metallic surfaces into a cumulative representation suitable for further analysis. It has been shown that the analysis of cumulative representations can effectively detect the presence of defects on metal surface images. A technique has been developed for converting an input image into a cumulative representation, which is then examined for abrupt transitions between different regions of the image by comparing the intensity changes in each row or column with the most frequent intensity value across the entire image. Datasets containing various images of metallic surface defects were selected and prepared. Software was developed to implement this image analysis method for defect detection. The software enables the construction of cumulative visualizations with adjustable input parameters, allowing researchers to evaluate the method’s effectiveness depending on the chosen visualization settings. It analyzes the resulting representation and, based on a specified deviation threshold, determines whether sharp intensity fluctuations are present. If such fluctuations are detected, the software reports the presence of defects on the original image; otherwise, the image is classified as defect-free. Additionally, the software can process entire sets of images and evaluate the accuracy of the method on a given dataset with selected parameters. Acceptable parameter ranges were determined, and testing was conducted across multiple datasets. The results indicate that the proposed algorithm achieves a high defect-detection accuracy, exceeding 80 percent correct classifications overall and up to 95 percent accuracy for images captured under uniform lighting conditions. The method was also found to be sufficiently flexible, allowing adaptation to different surface types and defect categories, making it suitable for a wide range of applications in metallurgical, automotive, aerospace, and other industrial sectors.