利用區域成長演算法進行多晶矽太陽能面板晶粒分析

科技高度發展的時代，人們對能源的依賴程度越來越高。因此尋找替代能源為當今重要課題，太陽能是未來重要的發電來源。太陽能電池主要分成單晶矽、多晶矽與非晶矽三大類，其中單晶矽的轉換率最高，其次是多晶矽，非晶矽轉換效率最低。但多晶矽太陽能電池製造成本較低且製程較簡單，成為市場上的主流。多晶矽面板平均晶粒尺寸愈大、晶界越少，其品質與光電轉換效率越佳，本論文提出簡單非破壞性測量，利用機器視覺對於多晶矽太陽能面板進行影像處理，分割出不同質性的晶粒，並利用不同質性的晶粒尺寸，計算平均晶粒尺寸作為多晶矽太陽能面板品質的衡量指標。首先應用影像強化和中值濾波的前處理。而後利用開發出的自動自適應性閥值分割演算法。我們透過該方法分割多晶矽太陽能面板影像，分割後封閉區域代表晶粒進行尺寸以及數量計算。最後將影像分解成四子圖像區域，利用管制圖方法估計每個子圖像中晶粒尺寸相對全距之母體標準差，再由相對全距的母體標準差分配每個子圖像所屬之權重。接著透過不同晶粒尺寸的數量和不同晶粒尺寸的面積，由這兩種指標結合區域的權重，估計太陽能面板的平均晶粒尺寸。利用平均晶粒尺寸結果作為指標，評估多晶矽太陽能面板的品質與轉換效率量測。

關鍵字

多晶矽太陽能電池；機器視覺；影像分割；中值濾波；區域成長演算法；自動自適應性閥值區域成長演算法；多晶矽太陽能電池轉換效率

並列摘要

In the era of high-technology, the degree of dependence on energy source becomes increasingly higher than before. The energy allocation problem influences not only in economic development, but also seriously affects the environment and ecology. Solar power is an attractive alternative source of energy. The multi-crystalline solar cell is widely accepted in the market because of its cheaper manufacturing cost. Multi-crystalline solar wafers have the characteristic that the larger average grain size surface, the less grain boundaries bring about higher quality and conversion efficiency. In this thesis, we would like to propose a new image processing method to assess the wafer quality that can segment separate grains in the solar wafer surface to calculate the number of grain and the average grain size as the evaluation indicators. The pre-processing steps based on the image contrast enhancement with a median filter are applied first, which has the advantage of enhancing image contrast and reducing noise while preserving the contour image. An automatic adaptive segmentation algorithm is developed based on the region growing segmentation algorithm, while retaining the fast convergence and a robust regional growth. The closed regions are obtained by using the region growing segmentation to segment the different grains, and then the grains sizes and numbers can be calculated. The image must be processed in the following two steps. First, we have to break down the image into 4 sub-images, and then use the control chart method to compute the standard deviation of the relative range of grain sizes to be assigned as the weights in each decomposed sub-image. Second, we proceed to calculating the numbers of different grain sizes and the areas of different grain sizes. In terms of the previously weights given in every sub-images, these two types of grain measures are used to estimate the average grain size. The resulting average grain size that represents the quality of solar wafers in crystallization can be adopted as a viable indicator of conversion efficiency.

並列關鍵字

Multi-crystalline Solar Cell ； Machine Vision ； Image Segmentation ； Median filter ； Region-Growing Algorithm ； Automatic adaptive Region-Growing approach ； Quality of solar wafer

參考文獻

1. 潘維治，Region Growing法於手寫數字辨識上的應用，碩士論文，私立大同大學資訊工程研究所，臺北，2004。

30. R. C. Gonzalez and R. E. Woods, Digital Image Processing, Prentice Hall: Pearson, 2008, pp. 311-334.

3. A. K. Joginipelly, Implementation of Separable & Steerable Gaussian Smoothers on an FPGA, Master Thesis, University of New Orleans, U. S. A., 2007.

8. D. Comaniciu and P. Meer, "Mean Shift Analysis and Applications," Proc. Seventh Int’l Conf. Computer Vision, 1999, pp.1197-1203.

9. D. Comaniciu and P. Meer, "Mean Shift: A Robust Approach toward Feature Space Analysis," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 24, no. 5, 2002, pp. 603-619.

國際替代計量

利用區域成長演算法進行多晶矽太陽能面板晶粒分析

全文下載

主題瀏覽