Principal Component Analysis Based on Wavelet Characteristics Applied to Automated Surface Defect Inspection

Automated visual inspection, a crucial manufacturing step, has been replacing the more time-consuming and less accurate human inspection. This research explores automated visual inspection of surface defects in a light-emitting diode (LED) chip. Commonly found on chip surface are water-spot blemishes which impair the appearance and functionality of LEDs. Automated inspection of water-spot defects is difficult because they have a semi-opaque appearance and a low intensity contrast with the rough exterior of the LED chip. Moreover, the defect may fall across two different background textures, which further increases detection difficulties. The one-level Haar wavelet transform is first used to decompose a chip image and extract four wavelet characteristics. Then, wavelet-based principal component analysis (WPCA) and Hotelling statistic (WHS) approaches are respectively applied to integrate the multiple wavelet characteristics. Finally, the principal component analysis of WPCA and the Hotelling control limit of WHS individually judge the existence of defects. Experimental results show that the proposed WPCA method achieves detection rates of above 93.8% and false alarm rates of below 3.6%, and outperforms other methods. A valid computer-aided visual defect inspection system is contributed to help meet the quality control needs of LED chip manufacturers. Key-Words: Surface defect inspection, Wavelet characteristics, Principal component analysis, Hotelling statistic, Machine vision system.