Learning graph-Fourier spectra of textured surface images for defect localization
Authors:
Tapan Ganatma Nakkina,
Adithyaa Karthikeyan,
Yuhao Zhong,
Ceyhun Eksin,
Satish T. S. Bukkapatnam
Abstract:
In the realm of industrial manufacturing, product inspection remains a significant bottleneck, with only a small fraction of manufactured items undergoing inspection for surface defects. Advances in imaging systems and AI can allow automated full inspection of manufactured surfaces. However, even the most contemporary imaging and machine learning methods perform poorly for detecting defects in ima…
▽ More
In the realm of industrial manufacturing, product inspection remains a significant bottleneck, with only a small fraction of manufactured items undergoing inspection for surface defects. Advances in imaging systems and AI can allow automated full inspection of manufactured surfaces. However, even the most contemporary imaging and machine learning methods perform poorly for detecting defects in images with highly textured backgrounds, that stem from diverse manufacturing processes. This paper introduces an approach based on graph Fourier analysis to automatically identify defective images, as well as crucial graph Fourier coefficients that inform the defects in images amidst highly textured backgrounds. The approach capitalizes on the ability of graph representations to capture the complex dynamics inherent in high-dimensional data, preserving crucial locality properties in a lower dimensional space. A convolutional neural network model (1D-CNN) was trained with the coefficients of the graph Fourier transform of the images as the input to identify, with classification accuracy of 99.4%, if the image contains a defect. An explainable AI method using SHAP (SHapley Additive exPlanations) was used to further analyze the trained 1D-CNN model to discern important spectral coefficients for each image. This approach sheds light on the crucial contribution of low-frequency graph eigen waveforms to precisely localize surface defects in images, thereby advancing the realization of zero-defect manufacturing.
△ Less
Submitted 1 December, 2023; v1 submitted 25 November, 2023;
originally announced November 2023.
