太陽能電池板表面瑕疵檢測; Surface Defect Detection for Solar Cells

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/9449

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/9449

Title:	太陽能電池板表面瑕疵檢測;Surface Defect Detection for Solar Cells
Authors:	陳心怡;Shin-Yi Chen
Contributors:	資訊工程研究所
Date:	2007-06-27
Issue Date:	2009-09-22 11:47:59 (UTC+8)
Publisher:	國立中央大學圖書館
Abstract:	隨著能源的日益耗盡，我們迫切地追求新能源的開發。太陽能即是一個具有潛在性的再生能源。而太陽能電池板是用來轉換太陽能的主要裝置。為了維持太陽能電池板的品質，我們發展了數種方法來檢測太陽能電池板的表面瑕疵。我們所檢測的對象包含數種不同種類的太陽能電池板。針對具有均勻表面的太陽能電池板，我們採用多選擇性的自動門檻值選取法來檢測瑕疵。我們額外地以追蹤的方法來檢測電池板的邊緣及線路部份。針對具有隨機紋理表面的太陽能電池板，若隨機紋理之間的對比並不是很明顯，我們採用以線偵測為基礎的方法，檢測出較亮的瑕疵；否則，我們先以非等方性的擴散方法模糊複雜的紋理背景，同時保留住較亮的瑕疵。接著仍然以線偵測為基礎的方法擷取出瑕疵。針對具有規律紋理表面的太陽能電池板，我們使用紋理學習的方法檢測瑕疵。我們事先訓練無瑕疵影像上的十種不同紋理。接著，將待測影像上的每一個小區塊和各種紋理模板做比較。若得到一個不相配的結果，表示此區塊內有瑕疵。我們以2048×2048 或者 640×640 解析度的單、多晶太陽能電池板影像來測試我們的偵測效能。從實驗結果顯示，我們所提出的方法已接近實用的階段。 Due to the decrease of the energy sources, the development of new energy sources is urgently pursued. One of the potential energy is the solar power. The solar cells are the primary devices to capture the solar energy. To maintain the quality of solar cells, we develop several inspection methods to detect defects on the solar cells. Several different kinds of solar cells are inspected. For the solar cell with uniform surface, we detect the defects using a multi-cased thresholding method. The border and busbar areas of solar cells are separated and inspected with a tracking style. For the solar cells with random texture surface, we use line-based detecting method to find brighter defects if the random textures of background is not clear; otherwise, we use an anisotropic diffusion based detection to blur the complex background while brighter defects would be preserved. Then the defects are also extracted by the line-based method. For the solar cells with regular texture surface, we use texture based method to detect the defects. We trained ten different texture patterns on defect-free solar images in advance. Then, we compare each patch of testing images with the corresponding trained patterns. The un-matched result shows defects in the patch. In the experiments, three proposed methods are evaluated with single crystal and polysilicon solar cells of 2048×2048 and 640×640 resolutions. The results show that the proposed methods are near practice for industry applications.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Size	Format

社群 sharing

Loading...