太陽能電池晶片表面自動化瑕疵檢測

Automatic surface defect inspection in multicrystalline solar wafers

李韋辰(Wei-Chen Li)

指導教授 ： 蔡篤銘

元智大學/工程學院/工業工程與管理學系/博士(2012年)

https://doi.org/10.6838/YZU.2012.00007

摘要

近年來由於原油價格不斷的上升且全球暖化程度日趨嚴重，促使各國不斷尋找替代能源，太陽能成為不可取代的發電來源；在太陽能產業上由於多晶矽太陽能電池(Solar cell)製造成本最為經濟，使得其成為市面上太陽能電池主流。太陽能晶片(Solar wafer)的表面若出現瑕疵將使下游太陽能電池之製造良率下降且使電池之發電效率不佳，因此本論文提出三種應用於多晶矽太陽能晶片之表面瑕疵檢測技術，可針對具有低對比且光源不均之無紋路背景進行瑕疵檢測，以及將模糊難辨之瑕疵自多晶太陽能晶片之異質性紋路表面區分出來。

本論文所提之前兩種太陽能電池/晶片瑕疵檢測法，使用霍夫轉換(Hough transform)為基之一維線段偵測演算法，此演算法首先利用水平與垂直灰階掃描線之資料點估計出直線，當任一偵測點偏離所估計之直線，則該點將被判定為瑕疵點。本論文的第一種瑕疵偵測法針對太陽能電池背面之低對比且光源不均之特性所造成的非穩定一維訊號，提出以霍夫轉換為基之直線偵測法，可有效的將低對比瑕疵偵測出來。第二種瑕疵偵測方法是針對多晶矽太陽能晶片之異質性紋路背景中的鋸痕瑕疵，利用傅立葉轉換(Fourier transform)影像重建技術將複雜晶格背景去除，使去除晶格後的影像成為光源不均之無紋路背景，便可結合本論文提出直線偵測法檢測出鋸痕瑕疵。前述兩種方法只針對特定瑕疵，而第三種瑕疵偵測方法是同時針對太陽能晶片上模糊難辨的指紋、髒污、鋸痕瑕疵，利用小波轉換(Wavelet transform)，透過萃取太陽能晶片影像上連續分解階層之小波能量值作為特徵，且將階層間之能量差異作為權重，建立足以分離正常晶格背景以及瑕疵之指標，可有效偵測瑕疵。

實驗結果顯示本論文所提出的直線偵測方法除了可有效偵測太陽能電池背面之凸塊瑕疵以及太陽能晶片上的鋸痕瑕疵之外，亦可應用於光源不均之影像上偵測出低對比瑕疵。本論文提出之第三種偵測方法可自複雜的多晶太陽能晶片背景晶格中，有效地分離出指紋、髒污、鋸痕等多種具模糊邊緣特性的瑕疵。

關鍵字

機器視覺 ； 瑕疵檢測 ； 多晶矽太陽能晶片 ； 霍夫轉換 ； 傅立葉轉換 ； 小波轉換

並列摘要

Solar power is an attractive alternative source of electricity. Multicrystalline solar cells dominate the market share owing to their lower manufacturing and material costs. The surface quality of solar wafer critically determines the conversion efficiency of the solar cell. In this research, three surface defect inspection techniques are presented for identifying low-contrast defects in non-textured surfaces (for backside solar cells) and detecting small local defects in inhomogeneous surfaces (for solar wafers).

The first two solar cells/wafers surface inspection algorithms use a Hough-like line detection method to identify defect points on 1D gray-level profiles of scan lines in the image. The conventional Hough transform requires a sufficient number of points lying exactly on the same straight line at a given parameter resolution so that the accumulator will show a distinct peak in the parameter space. It fails to detect a line in a non-stationary signal. The first proposed Hough-like algorithm can effectively detect the low-contrast defects in the unevenly-illuminated surface of a backside solar cell. In the second proposed method, the inhomogeneous background of multicrystalline grains in a solar wafer image can be effectively removed by properly selecting the band-rejection region in the Fourier spectrum, and then the proposed Hough-like line detection technique is used to identify saw-mark defects in a solar wafer. The third surface defect inspection method is based on the two-dimensional discrete wavelet transform, and is applied to the detection of various defect types. It takes the energy difference between two consecutive decomposition levels as a clue to enhance the discriminant features extracted in individual decomposition levels and generates a better discriminant measure for identifying defects with scattering and blurred edges. Experimental results have shown that the proposed methods perform effectively for detecting low-contrast bump in the unevenly-illuminated backside solar cell, and various defects of stain, saw-mark, fingerprint and contaminant in the inhomogeneous solar wafer surface

並列關鍵字

Multicrystalline solar wafer ； Hough transform ； Fourier transform ； Wavelet transform ； Machine vision ； Surface inspection ； Defect detection

參考文獻

Rau J. Y. and L. C. Chen, 2003, Fast straight line detection using Hough transform with principal axis analysis, Journal of Photogrammetry and Remote Sensing, vol. 8, pp. 15-34.

Abdesselam A., 2010, A multi-resolution texture image retrieval using fast Fourier transform, Journal of Engineering Research, vol. 7, pp. 48-58.

Antonini M., M Barlaud, P. Mathieu and I. Daubechies, 1992, Image coding using wavelet transform, IEEE Trans. Image Process., vol. 1, pp. 205-220.

Arivazhangan S. and L. Ganesan, 2003, Texture classification using wavelet transform, Pattern Recognition Letters, vol. 24, pp. 1513-1521.

Bergen J. R. and H. Shvaytser, 1991, A probabilistic algorithm for computing Hough transforms, Journal of Algorithms, vol. 12, pp. 639-656.