Classification of biscuit tiles for defect detection using Fourier transform features

Automated quality inspection remains a challenge in ceramic tile production. While computer vision systems have been explored, their effectiveness on textured surfaces is limited. Traditional methods struggle with the complex patterns of biscuit tiles, which are unglazed and pre-fired. This gap motivated the search for better feature extraction techniques. Prior research has shown that Fourier-based methods can capture texture patterns effectively. However, no prior work had resolved the issue of applying these methods to biscuit tiles specifically. The need to reduce energy and material waste in tile manufacturing drove this investigation. This study addresses the lack of robust defect detection methods for early-stage ceramic products.

The goal was to develop a reliable method for identifying defects in biscuit tiles before firing. Biscuit tiles are intermediate products that require costly energy for firing, so early defect detection could save resources. The study aimed to improve on existing feature extraction techniques by using Fourier transform features. The approach focuses on surface texture analysis rather than post-firing inspection. The researchers propose a novel method based on Fourier spectrum annuli. This method was tested on real-world ceramic tile datasets. The study sought to determine if this new approach could outperform standard methods. The ultimate aim was to provide a practical solution for manufacturers.

The proposed method uses Fourier spectrum annuli to extract features from biscuit tile images. The process begins by capturing high-resolution images of the tile surfaces. These images are then transformed into their Fourier spectrum representations. The method focuses on concentric annuli within the Fourier domain to capture texture patterns. Each annulus is analyzed for energy distribution and spatial frequency characteristics. The extracted features are fed into a machine learning classifier for defect detection. The system was trained and validated using real-world ceramic tile datasets. Performance metrics such as F1 score were used to evaluate the method's effectiveness.

The Fourier spectrum annuli method achieved an F1 score of 0.9236 on the Black Random Stripes tile dataset. On the Stripes Brown Light dataset, the method reached an F1 score of 0.8866. These results outperformed several established feature extraction techniques. The method demonstrated high accuracy in distinguishing defective from non-defective tiles. The Fourier-based approach captured texture variations more effectively than traditional methods. The results suggest that this method is well-suited for textured surfaces like biscuit tiles. The performance metrics indicate strong potential for industrial application. The study provides evidence that Fourier spectrum annuli can enhance defect detection accuracy.

The authors propose that the Fourier spectrum annuli method offers a practical solution for biscuit tile inspection. The results suggest that this approach can reliably detect defects before firing. The method's performance on real-world datasets supports its potential for industrial use. The study highlights the importance of texture-based feature extraction in ceramic quality control. The authors emphasize that early defect detection can reduce manufacturing costs. The findings indicate that Fourier transform features are effective for textured surfaces. The study does not claim that this is the only viable method, but it suggests that it is a strong candidate. The authors propose that this method could be integrated into existing quality control systems.