機器視覺應用於太陽電池之表面瑕疵檢測

Surface Defects Detection of Solar Cell by Using Machine Vision

郭冠志(Kuan-Chih Kuo)

指導教授 ： 吳明川

國立臺北科技大學/機電學院/機電整合研究所/碩士(2007年)

https://doi.org/10.6841/NTUT.2007.00142

摘要

在太陽電池(Solar Cell)的品管檢測中，可分為電信檢測以及表面瑕疵檢測，而本研究針對太陽電池的表面可見瑕疵，包括表面髒污以及佈線不良這兩種瑕疵，利用電腦影像處理技巧來進行檢測。本研究所檢測之太陽電池表面具有纹路之特徵，而在太陽電池表面之黑色細節灰階值並非呈現一致之特性，且在太陽電池製造過程經過修角，以致太陽電池為一類似八角型之結構，造成四角落檢測時容易因背景而導致誤判情況。也由於具有許多的線條紋路，傳統上以人工方式進行檢測，不僅容易誤判，且效率低落無法達成固定之品質標準，有鑑於此，本研究期能發展一套檢測技術，進而達到自動化之發展。
研究中所使用的兩種檢測方式，(1)小波轉換(Wavelet Transform)檢測法：利用小波轉換影像轉換與還原技巧應用於太陽電池之檢測，實驗結果驗證現有文獻以同方式應用於LCD檢測也可應用於太陽電池之瑕疵檢測，不過在邊緣之影像以及纹路交接點容易出現誤判情況；(2)數學形態學(Mathematical Morphology)檢測法：完全以數學形態學之概念進行檢測，此方式必須先針對樣本先進行細部結構元素參數設定，建立好系統後進行檢測，可發現以本研究之實驗結果檢測正確率可達約90%，而整體檢測時間也以數學形態學法優於小波轉換法。

關鍵字

太陽電池 ； 小波轉換 ； 數學形態學

並列摘要

On the product quality inspection of solar cell, it can be divided into electrical test and surface defects detection. This study is mainly aimed at the visual defects detection by using digital image processing, including the surface pollution and electrical wiring incorrectness. There is a textured characteristic on the surface of solar cell, and it is not identical on grey-level values of the dark details. Solar Cell is analogous to an octagonal structure after the manufacturing of cutting four corners; hence, erroneous judgments maybe happened when inspecting four corners. Because of the textured feature, it might be failure to inspect defects by using human vision. It also can not reach the quality standard and get low efficiency. Therefore, this paper is to develop an inspection method to achieve the goal of automated inspection.
There are two methods in this paper. The first method is Wavelet Transform which is using Wavelet Transform on image transform and restoration to inspect solar cell. According to the experimental result, the method being used currently to inspect LCD can also be used in solar cell inspection; however it could be error on inspecting the image edge and texture intersection point. The second method is Mathematical Morphology which is using the conception of Mathematical Morphology completely. Setting the structure element first that based on the testing samples. After the system has been established, it is found that the inspection accuracy of this method is about 90%. In addition, the speed to execute Mathematical Morphology method is faster than Wavelet Transform method as well.

並列關鍵字

Solar Cell ； Wavelet Transform ； Mathematical Morphology

參考文獻

[28] 陳錦宗，小波分析LCD面板瑕疵之研究，碩士論文，私立元智大學電機工程系，桃園，2006。

[6] K. Nakashima, “Hybrid Inspection System for LCD Color Filter Panels,” Instrumentation and Measurement Technology Conference, vol.2, 1994, pp.689-692.

[8] 黃國唐，彩色紋路表面之瑕疵檢測，碩士論文，私立元智大學工業工程與管理研究所，桃園，1999。

[9] 楊榮華，應用灰階共變異矩陣之多重指標於瑕疵檢測，碩士論文，私立元智大學工業工程與管理研究所，桃園，2002。

[10] Fu Zhuang, Zhao Yanzheng, Liu Yang, Cao Qixin, Chen Mingbo, Zhang Jun, Jay Lee, “Solar cell crack inspection by image processing,” International Conference on the Business of Electronic Product Reliability and Liability, 2004, pp.74-80.