Mean Shift-Based Defect Detection in Multicrystalline Solar Wafer Surfaces

This paper has presented a mean shift-based machine vision method for detecting fingerprint and contamination defects in multicrystalline solar wafers. The defect types involve random gradient directions. whereas the normal grain edges generally present more consistent gradient directions in a small spatial window. The entropy of gradient directions is then used as the range feature. The pixel coordinates along with the entropy form the feature space of the image. The mean-shift smoothing can effectively remove noise and residuals of crystal grain edges and preserve only the defective pixels in the filtered image. A simple adaptive threshold can thus be used to segment the defective region in the filtered entropy image.

About core

Check Also

Ukraine Eyes Billions in Euros from Europe Electricity Exports

Ukraine hopes to make €1.5 billion from electricity exports to the European Union by the …

Leave a Reply

Your email address will not be published.