201250234 六、發明說明： 【發明所屬之技術領域】 本發明關於可用於太陽能電池内部缺陷之電腦程式產 品、偵測設備及偵測方法，特別是關於影像處理及自動偵 測技術。 【先前技術】 世界文明隨著科技蓬勃發展，各種能源紛紛被發掘採 用，其中又以石油為首，近年來隨著世界人口暴漲，石油 的存量快速減退，太陽能成了許多國家能源科技發展重點 項目之一，並且提供了相當多的經費來補助此一產業。太 陽能電池可以直接將光能轉換為直流電，其特性是無需燃 料、無廢棄物與汙染、無轉動組件與噪音，而且外型、尺 寸可以任意變化，使用壽命可長達20年，所以很適合消費 性產品或是一般家庭使用。 因應如此新能源之潮流，本國亦有許多廠商投入太陽 產業期中，由於本國廠商大多是以製造為主，因此如何有 效提升生產效能、降低生產成本便成了一個重要的課題。 其中，可想而知太陽能電池之良率測試即是非常重要的技 術重點。 目前許多檢測的研究都是針對太陽能電池外部瑕疵利 用攝影機取像，然後運用影像處理方式自動進行檢測或是 分級，然而，太陽能電池除了肉眼可見之外部瑕疵外，另 外有一些内部瑕疵是來自於電池内部微裂所造成的瑕疵。 由於這些内部瑕疵發生在太陽能板内部，因此無法以肉眼 201250234 直接判定，同時，現有自動檢測系統也無法檢測出來。但 由於内部微裂的内部瑕疵會造成整個太陽能板發電效率降 低，並且可能在產品使用一段時間之後造成面板斷裂，所 以事實上是必須解決的技術問題。 【發明内容】 因此，本發明之一態樣是在提供用於太陽能電池内部 缺陷的影像處理之電腦程式產品、偵測設備及偵測方法。 依據本發明一實施方式，一種影像處理之電腦程式產 品，用以偵測太陽能電池内部缺陷，且至少執行下列步驟： 讀入步驟、第一到第四處理步驟及輸出步驟。讀入步驟讀 入影像，影像包含複數個像素。第一處理步驟利用最大臨 界值與最小臨界值來與像素進行比較以濾除影像之複數個 雜訊像素而產生複數個候選像素。第二處理步驟將候選像 素與内框臨界值比較以辨認出候選像素中之複數個内框像 素。第三處理步驟將内框像素之強度設為大於缺陷臨界 值。第四處理步驟將包含内框像素之候選像素依序與缺陷 臨界值比較以找出複數個缺陷像素。而輸出步驟輸出該缺 陷像素之對應位置。 依據上述實施方式，此電腦程式產品更可以在第二處 理步驟前，執行斜率計算步驟，其尋找由候選像素組成之 一區域的邊緣四點，以計算區域之斜率。上述電腦程式產 品也可以即時輸出影像之資料給一顯示介面。而所謂的雜 訊像素可包含複數個背景像素。 依據本發明另一實施方式，一種執行如前述之電腦程 5 201250234 太陽能電池内部缺陷之偵測設備也被提出。此偵 =備I含電源供應器、電#麵合攝影機、電腦主機、如 “2式產品、顯示裝置°而電源供應器供應直流 j陽能電池’使太陽能電池發出各種波長的光譜。 攝影機包含濾除可見光進入電荷耦合攝影機之濾 昭、配置以對向太陽能電池來攝取太陽能電池的光譜 光說昭兄碩。電腦主機為電性連接電荷輕合攝影機以接收 像並轉成f彡像，且執行前述電腦程式產品以進行影 對應位置顯示裝置與電腦主機電性連接並顯示缺陷像素之 之偵！:據本發明又一實施方式，一種太陽能電池内部缺陷 像+Z、1方法亦被提出。此偵測方法至少包含偏壓步驟、攝 路、轉存步驟及影像處理步驟。偏壓步驟係指偏壓電 步驟ft壓給太陽能電池，使太陽能電池發出光譜。攝像 i光=指濾片鏡頭以過濾掉可見光之方式取得太陽能電池 數^^。轉存步驟係指運算單元儲存光譜照片為由複 大臨象素組成之影像。影像處理步驟係指運算單元利用最 ^界值與最小臨界值來與像素進行比較以濾除影像之 ^雜訊像素而產线數個㈣像素，運算單元再將 傻去與内^臨界值比較以辨認出候選像素中之複數個内框 、，運算單元又將内框像素之強度設為大於缺 及運算單元最後將包含賴像素之㈣像素依序與缺陷 α界值比較以找出複數個缺陷像素。 、 、依據上述實施方式，此偵測方法可包含在運算 生複數個候選像素後’尋找雜選像素之邊緣四點，7^產 Μ舌十 6 201250234 算候選像素間之斜率之步驟。偵測方法亦可包含於顯示介 面即時輸出影像之步驟或包含輸出缺陷像素之對應位置之 步驟。前述之雜訊像素包含複數個背景像素。 綜上所述，利用本發明所提出之用於太陽能電池内部 缺陷的影像處理之電腦程式產品、偵測設備及偵測方法可 以輕易在生產的過程中，檢測出内部瑕疵，並且進一步根 據内部瑕疵所發生的位置，改善其產線的製程。如此一來， 便可以有效率地提升太陽能電池製造的品質與良率。 【實施方式】 本發明提出之用於太陽能電池内部缺陷的影像處理之 電腦程式產品、偵測設備及偵測方法可用以偵測如第1圖 所示之太陽能電池的内部缺陷，第1圖中可以得知太陽能 電池100包含2x9個太陽能電池單元110，而各太陽能電 池單元110間被内框120區隔。 依據本發明之一實施方式所提出之偵測設備包含電源 供應器（習知之直流電壓供應器）、電荷耦合攝影機、電腦 主機、影像處理之電腦程式產品、顯示裝置（如電腦螢幕）。 而電源供應器供應直流偏壓給太陽能電池，使太陽能電池 發出各種波長的光譜。電荷耦合攝影機包含濾除可見光進 入電荷耦合攝影機之濾光透鏡（請參考第2圖，波型圖顯 示其僅能讓波長800nm以上之光譜通過）與配置以對向太 陽能電池來攝取太陽能電池的光譜照片的鏡頭（如16 bits 灰階CCD camera，可以解析度765x510攝像），為了可以 清楚攝像到強度相較之下偏弱之太陽能電池之光譜照片， 201250234 曝光時間至少需30秒以上。 接著’電腦主機為電性遠桩雷 譜照，轉成料，聽===== ^置和裝置與％腦主機電性連接並顯示缺陷像素之對應 凊參考第3圖可以得知上述之雷腦 處理之_㈣雜，詳纟请B對於影像 步驟310 :讀入步驟讀入影像，影像包含複數個像素。 步驟320:第-處理步驟利用最大臨界值與最小臨界 值來與像素進行比較以滤除影像之複數個雜訊 ^ 複數個候選像素。 〃 步驟330 :第二處理步驟將候選像素與内框臨界值比 較以辨認出候選像素中之複數個内框像素。 缺陷=值補：第三處理步驟將内框像素之強度設為大於 步驟350:第四處理步驟將包含内框像素之 依序與缺陷臨界值比較以找出複數個缺陷像素。、、、 步驟360 :輸出步驟輸出該缺陷像素之對應位置。 在拟更進一步細言之’步驟320 (雙臨界值雜訊減除法） 係對於影像中的每一個像素，選取一個ΝχΝ大】 對於遮罩中每一像素的輝度值，用最小臨界值丁;盥：罩: = Τ2(Τ1<Τ2)與之比較’若遮革中大部分像素 ^接近最小臨界值T1，=m像素位於背景，因此將遮 軍中央像素的灰階值設為黑色；反之，若遮罩中大部分像 素的灰階值較接近最大臨界值T2 ’則判斷像素位於太陽能 201250234 電池因此保持遮罩巾央像素的灰階值*改變。詳細演算 法如下所示： A. 設定最小臨界值T1與最大臨界值了2，令τι<τ2 ; cntl=cnt2=〇 B. 對於遮罩内的每—個像素灰階值，計算其與^與 Τ2的差異絕對值&、们 C. 以⑴、犯大小為依據更新cntl與cnt2，若dl小於 d2則cntl加1 ;否則cm2加1 D. 以cntl與cnt2的值決定輸出像素值，若_ >⑶口 則遮罩内的中央像錢階值為黑色；㈣保持原值 不變 而經過步驟320之後，經移除之雜訊像素亦包含背景 ^素。也就是說’經過步驟32〇後所產生之候選像素會是 太陽能電池100對應之影像像素集合。 、若進—步細述步驟330〜340，此兩步驟之主要目的是 $ J·、辨識太陽能電池單元11〇拼湊處所形成的黑色内框， 就疋内框120之對應像素。而藉由步驟33〇〜34〇避免將 内樞誤判為内部瑕疵。步驟33〇〜34〇為選定一内框臨界值 T3，逐步掃描圖形中水平或垂直線，將T3與每一掃瞄像 素的灰階值做比較，判斷掃瞄線中是否有大量像素的灰階 值小於Τ3，若是，則將掃描軸中所有的像素點色彩變更為 白色；否則保留像素原來灰階值。詳細演算法如下所示： Α.設定内框臨界值Τ3與Κ Β.逐步抓取候選像素中水平軸每一個像素點的灰階 值x( i，j )與臨界值Τ3做比較’若χ( i，j )小於 9 201250234 Τ3，貝1j cnt3 加 1 C. 若cnt3 > K，則判定水平軸的像素點為内框像素， 因此將水平軸像素值設定為白色（以0〜255之尺度 來看，白色為255其強度勢必大於缺陷臨界值） D. 對於影像中的垂直軸重複(Β)與（C)，辨識出内框的 垂直線 若進一步細述步驟350 (區域成長法）係將包含内框 像素之候選像素依序與缺陷臨界值比較以找出複數個缺陷 像。詳細演算法如下所示： Α.令C為目前已經尋找過的内部缺陷區域，初始化 C=0201250234 VI. Description of the Invention: [Technical Field] The present invention relates to computer program products, detection devices and detection methods that can be used for internal defects of solar cells, particularly regarding image processing and automatic detection technology. [Prior Art] With the vigorous development of science and technology, various kinds of energy have been discovered and adopted, among which oil is the first. In recent years, with the world population soaring, the stock of oil has rapidly declined, and solar energy has become the key project of energy science and technology development in many countries. First, and provided a considerable amount of funds to subsidize this industry. Solar cells can directly convert light energy into direct current. Its characteristics are fuel-free, waste-free and pollution-free, no rotating components and noise, and its appearance and size can be changed arbitrarily. The service life can be as long as 20 years, so it is suitable for consumption. Sex products or general household use. In response to the trend of such new energy sources, many manufacturers in the country have invested in the solar industry. Since most of the domestic manufacturers are mainly manufacturing, how to effectively improve production efficiency and reduce production costs has become an important issue. Among them, it is conceivable that the yield test of solar cells is a very important technical focus. At present, many researches on the outside of the solar cell are taken by the camera, and then image processing is used to automatically detect or classify. However, in addition to the visible external sputum, the solar cell has some internal 瑕疵 from the battery. The flaw caused by internal micro-cracking. Since these internal flaws occur inside the solar panel, they cannot be directly judged by the naked eye 201250234, and the existing automatic detection system cannot detect it. However, due to the internal flaws in the internal micro-cracking, the efficiency of the entire solar panel is reduced, and the panel may be broken after a period of use of the product, so it is a technical problem that must be solved. SUMMARY OF THE INVENTION Accordingly, one aspect of the present invention is a computer program product, a detection device, and a detection method for providing image processing for internal defects of a solar cell. According to an embodiment of the invention, a computer program product for image processing is used for detecting internal defects of a solar cell, and at least performing the following steps: a reading step, first to fourth processing steps, and an output step. The reading step reads the image and the image contains a plurality of pixels. The first processing step uses the maximum critical value and the minimum critical value to compare with the pixels to filter out a plurality of noise pixels of the image to generate a plurality of candidate pixels. The second processing step compares the candidate pixels with the inner frame threshold to identify a plurality of inner frame pixels in the candidate pixels. The third processing step sets the intensity of the inner frame pixels to be greater than the defect threshold. The fourth processing step sequentially compares the candidate pixels including the inner frame pixels with the defect threshold to find a plurality of defective pixels. The output step outputs the corresponding position of the defective pixel. According to the above embodiment, the computer program product can further perform a slope calculation step for finding the edge four points of a region composed of candidate pixels to calculate the slope of the region before the second processing step. The above computer program product can also instantly output image data to a display interface. The so-called noise pixels can include a plurality of background pixels. According to another embodiment of the present invention, a detecting apparatus for performing internal defects of a solar cell as described above is also proposed. This detection=preparation I includes power supply, electric #face camera, computer mainframe, such as "2 type product, display device ° and power supply supply DC j solar battery" to enable solar cells to emit spectra of various wavelengths. Filtering the visible light into the charge-coupled camera, and configuring the spectrum light to capture the solar cell to the solar cell, the computer host is electrically connected to the charge-collecting camera to receive the image and convert it into an image. The computer program product is executed to electrically connect the image corresponding position display device to the computer host and display defective pixels. According to still another embodiment of the present invention, a solar cell internal defect image +Z, 1 method is also proposed. The detecting method includes at least a biasing step, a path of retreating, a step of transferring, and an image processing step. The biasing step refers to biasing the electric step to the solar cell to cause the solar cell to emit a spectrum. The imaging i-light refers to the filter lens. The number of solar cells is obtained by filtering out visible light. The transfer step refers to the operation unit storing the spectrum photo as a complex pixel. The image processing step refers to the operation unit using the most boundary value and the minimum threshold value to compare with the pixel to filter out the image noise pixels and generate a plurality of (four) pixels, and the operation unit will be stupid and internal. ^The threshold value is compared to identify a plurality of inner frames in the candidate pixels, and the operation unit sets the intensity of the inner frame pixels to be larger than the missing operation unit, and finally compares the (four) pixels of the included pixels with the defect α boundary value. Finding a plurality of defective pixels. According to the above embodiment, the detecting method may include searching for four edges of the candidate pixel after computing the plurality of candidate pixels, and calculating the edge of the candidate pixel. The step of detecting the slope may also include the step of outputting the image on the display interface or the step of outputting the corresponding position of the defective pixel. The aforementioned noise pixel includes a plurality of background pixels. The computer program product, detection device and detection method for image processing of internal defects of solar cells can be easily detected during production. Internally, and further improving the manufacturing process of the production line according to the position where the internal defects occur. Thus, the quality and yield of the solar cell manufacturing can be efficiently improved. The computer program product, the detecting device and the detecting method of the image processing of the internal defect of the battery can be used to detect the internal defects of the solar cell as shown in FIG. 1 , and it can be seen from FIG. 1 that the solar cell 100 includes 2×9 solar cells. The unit 110 is separated from each of the solar cells 110 by the inner frame 120. The detecting device according to an embodiment of the present invention comprises a power supply (a conventional DC voltage supply), a charge coupled camera, a computer host, A computer program product for image processing, a display device (such as a computer screen), and a power supply that supplies a DC bias voltage to the solar cell to cause the solar cell to emit spectra of various wavelengths. A charge-coupled camera consists of a filter lens that filters out visible light into a charge-coupled camera (see Figure 2, which shows that the spectrum can only pass through a wavelength above 800 nm) and the spectrum of the solar cell that is configured to oppose the solar cell. The lens of the photo (such as 16-bit grayscale CCD camera, can be resolved to 765x510 camera), in order to clearly capture the spectrum of the solar cell with weaker intensity, the exposure time of 201250234 is at least 30 seconds. Then, the computer mainframe is electrically far-reaching, and the material is converted into a material. The device is electrically connected to the % brain host and displays the corresponding pixels. Refer to Figure 3 for the above. Ray ray processing _ (four) miscellaneous, please refer to B for image step 310: read the step to read in the image, the image contains a plurality of pixels. Step 320: The first processing step uses the maximum threshold and the minimum threshold to compare with the pixels to filter out a plurality of noises of the image. Step 330: The second processing step compares the candidate pixels with the inner frame threshold to identify a plurality of inner frame pixels in the candidate pixels. Defect = Value Complement: The third processing step sets the intensity of the inner frame pixel to be greater than step 350. The fourth processing step compares the sequentially included pixels with the defect threshold to find a plurality of defective pixels. , , , Step 360: The output step outputs the corresponding position of the defective pixel. In the further step of 'Step 320 (Double Threshold Noise Reduction), for each pixel in the image, select a large value. For the luminance value of each pixel in the mask, use the minimum threshold value;盥: hood: = Τ2(Τ1<Τ2) Compare with it's 'most of the pixels in the occlusion are close to the minimum threshold T1, =m pixels are in the background, so the grayscale value of the occlusion central pixel is set to black; If the grayscale value of most of the pixels in the mask is closer to the maximum critical value T2', then the pixel is determined to be in the solar 201250234 battery, thus keeping the grayscale value* of the masking central pixel* changed. The detailed algorithm is as follows: A. Set the minimum threshold T1 and the maximum threshold to 2, let τι<τ2; cntl=cnt2=〇B. For each pixel grayscale value in the mask, calculate it and ^ Absolute value & 们2, C. (1), update cntl and cnt2 based on the size of the nucleus, if dl is less than d2 then cntl plus 1; otherwise cm2 plus 1 D. The value of cntl and cnt2 determines the output pixel value, If _ > (3) port, the central image money value in the mask is black; (4) after the original value remains unchanged, after step 320, the removed noise pixel also includes the background element. That is to say, the candidate pixels generated after the step 32 is the image pixel set corresponding to the solar cell 100. If the steps 330 to 340 are further described, the main purpose of the two steps is to identify the black inner frame formed by the solar cell unit 11 and the corresponding pixel of the inner frame 120. By step 33〇~34〇, the internal arm is mistakenly judged as internal 瑕疵. Steps 33〇~34〇 are selected to select an inner frame threshold T3, and gradually scan the horizontal or vertical lines in the graph, compare T3 with the grayscale value of each scan pixel, and determine whether there is a large number of pixels in the scan line. The value is less than Τ3, and if so, the color of all pixels in the scan axis is changed to white; otherwise, the original grayscale value of the pixel is retained. The detailed algorithm is as follows: Α Set the inner box threshold Τ3 and Κ Β. Gradually grab the gray scale value x( i,j ) of each pixel in the horizontal axis of the candidate pixel and compare it with the critical value Τ3. ( i, j ) is less than 9 201250234 Τ 3, Bay 1j cnt3 plus 1 C. If cnt3 > K, it is determined that the pixel of the horizontal axis is the inner frame pixel, so the horizontal axis pixel value is set to white (with 0 to 255 In terms of scale, the white is 255 and its intensity is bound to be greater than the defect threshold. D. For the vertical axis in the image, repeat (Β) and (C), identify the vertical line of the inner frame. If you further detail step 350 (regional growth method) The candidate pixels including the inner frame pixels are sequentially compared with the defect threshold to find a plurality of defect images. The detailed algorithm is as follows: Α. Let C be the internal defect area that has been searched for, initialize C=0
B. 依序尋找強度小於缺陷臨界值Θ且尚未被分類至 任何存在的内部缺陷區域的像素；若無此種像素， 則離開；反之，則令C=C+1，且歸類此像素在内部 缺陷區域C，又將此像素加入一空白集合SB. sequentially look for pixels whose intensity is less than the defect threshold and have not been classified into any existing internal defect regions; if there is no such pixel, leave; otherwise, let C=C+1, and classify the pixel Internal defect area C, this pixel is added to a blank set S
C. 尋找S中相鄰像素p。若p之強度也小於缺陷臨界 值0且尚未被分類至任何存在的内部缺陷區域的 話，將p歸類在内部缺陷區域C，又將p鄰近之像 素加入S D. 重複(C)以找到所有屬於S的像素p E. 回到（B) 另一方面，當步驟310讀入之影像與視窗非平行時， 或說其影像非為正向而為斜向時，電腦程式產品在步驟320 之後，步驟330之前，會進行步驟325，也就是執行斜率 計算步驟，其尋找由該候選像素組成之一區域的邊緣四 201250234 二==區域之斜率。藉此協助找出與區域平行之内 時輸理的過程當中，更可以即 二广7圖之實:例 ^譜照片經過步驟㈣〜之影像 實 例2即代表第】圖之太 圖之實知 31〇〜360(含步驟32〇旦 之斜向光譜照片經過步驟 更可以增加即時顯干各H象。並且在即時顯示的視窗中， 應用設備之需求 像素之強度/座標等功能以符合各 實施例1 ，參考第4圖，可以得知經過曝光相，内部、& 成太％能電池發光中光譜散出的效能不彰，因此影像二 陰影的出現’顯示出裂痕的位置 有 沒有辦法直錢行__作。帛=== 同時也可以把太陽電池在影像中的位置觸楚;;去除雜讯， =2 = 2 33: 34。之影像’可以看_，内框 ，又疋為白色之後，内框與内部瑕庇之差異十 、.坐可以避免下-步驟的誤判情形。最後，第7圖係 =步驟350之影像，無庸置疑地太陽能電池内部= 位置已經被有效地找出。 、心的 實施例2 、，參考第8圖’可以得知經過曝光時間，内 成太陽能電池發光中光譜散出的效能不彰，因此影^ = 201250234 陰影的出現，顯示出裂痕的位置，而雜訊的干擾十分明顯， 沒有辦法直接進行偵測的動作，並且由於太陽能電池沒有 與鏡頭校正座標，造成影像歪斜。第9圖係第8圖經過步 驟320、325、330、340之影像，便可以得知雜訊已經被濾 除，且内框也已辨識設定完畢，可以進行下一步驟。最後， 第10圖係第8圖經過步驟350之影像，無庸置疑地太陽能 電池内部缺陷的位置已經被有效地找出。 由上述本發明實施方式可知，應用本發明所提出之用 於太陽能電池内部缺陷的影像處理之電腦程式產品、偵測 設備及偵測方法可以輕易在生產的過程中，檢測出内部瑕 疵，並且進一步根據内部瑕疵所發生的位置，改善其產線 的製程。如此一來，便可以有效率地提升太陽能電池製造 的品質與良率。 雖然本發明已以實施方式揭露如上，然其並非用以限 定本發明，任何熟習此技藝者，在不脫離本發明之精神和 範圍内，當可作各種之更動與潤飾，因此本發明之保護範 圍當視後附之申請專利範圍所界定者為準。 【圖式簡單說明】 為讓本發明之上述和其他目的、特徵、優點與實施例 能更明顯易懂，所附圖式之說明如下： 第1圖是習知的太陽能電池之結構示意圖。 第2圖本發明一實施方式中使用的濾片的濾光波型 圖。 第3圖係繪示依照本發明之一實施方式的一種電腦程 12 201250234 式產品的執行步驟流程圖。 第4圖係第1圖之太陽能電池之正向光譜照片（步驟 310)。 第5圖係第4圖經過步驟320之影像。 第6圖係第4圖經過步驟33〇、34〇之影像。 第7圖係第4圖經過步驟350之影像。 第8圖係第1圖之太陽能電池之斜向光譜照片（步驟 310)。 第9圖係第8圖經過步驟32〇、325、33〇、34〇之影像。 第10圖係第8圖經過步驟350之影像。 【主要元件符號說明】 10〇 :太陽能電池 110 :太陽能電池單元 120 :内框 310〜360 :步驟 13C. Find the adjacent pixel p in S. If the strength of p is also less than the defect threshold of 0 and has not been classified to any existing internal defect area, p is classified into the internal defect area C, and p adjacent pixels are added to S D. Repeat (C) to find all Pixel belonging to S. Back to (B) On the other hand, when the image read in step 310 is not parallel to the window, or when the image is not positive and oblique, the computer program product is after step 320. Before step 330, step 325 is performed, that is, a slope calculation step is performed, which searches for the slope of the edge 4 201250234 2 == region of a region composed of the candidate pixels. In this process, when assisting in finding out the process of paralleling with the area, it can be the same as the picture of the second and the wide picture: the case of the picture ^ after the step (4) ~ the image example 2 represents the figure of the figure 31〇~360 (including the oblique spectrum photo of step 32), the step can further increase the H-image instantly, and in the instant display window, the application device needs the intensity/coordinate of the pixel to conform to the implementation. Example 1, referring to Figure 4, it can be seen that after the exposure phase, the internal, &% of the energy of the battery emits light in the spectrum is not effective, so the appearance of the image two shadows 'shows the location of the crack is not straight Money line __作.帛=== At the same time, you can also touch the position of the solar cell in the image; remove the noise, =2 = 2 33: 34. The image 'can see _, the inner frame, and then After white, the difference between the inner frame and the inner shelter is ten. Sitting can avoid the misjudgment of the next step. Finally, the image of step 7 = step 350, undoubtedly the internal solar cell = position has been effectively found , heart embodiment 2, reference 8 picture 'can be known that after the exposure time, the performance of the spectrum emission in the solar cell illuminating is not good, so the shadow ^ = 201250234 the appearance of the shadow, showing the location of the crack, and the interference of the noise is very obvious, there is no way directly The detection action is performed, and since the solar cell does not correct the coordinates with the lens, the image is skewed. Fig. 9 is the image of steps 320, 325, 330, and 340, and it can be known that the noise has been filtered out. And the inner frame has been identified and set, and the next step can be performed. Finally, the image of step 10 in Fig. 10 is passed through the image of step 350, and it is undoubted that the position of the internal defect of the solar cell has been effectively found. It can be seen that the computer program product, the detecting device and the detecting method for image processing of the internal defects of the solar cell proposed by the present invention can easily detect the internal flaw during the production process, and further according to the internal environment. The location where it takes place, improving the process of its production line, so that solar cell manufacturing can be efficiently improved The present invention has been disclosed in the above embodiments, but it is not intended to limit the invention, and various modifications and refinements can be made without departing from the spirit and scope of the invention. Therefore, the scope of the present invention is defined by the scope of the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features, advantages and embodiments of the present invention will become more apparent. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a schematic view showing the structure of a conventional solar cell. Fig. 2 is a view showing a filter pattern of a filter used in an embodiment of the present invention. Fig. 3 is a view showing one of the present invention. A flow chart of the execution steps of a computer program 12 201250234 product of the embodiment. Figure 4 is a photograph of the forward spectrum of the solar cell of Figure 1 (step 310). Figure 5 is an image of the fourth step through step 320. Figure 6 is a picture of Figure 4 through steps 33, 34. Figure 7 is an image of the fourth step through step 350. Fig. 8 is a diagonal spectrum photograph of the solar cell of Fig. 1 (step 310). Figure 9 is an image of Figure 8 through steps 32, 325, 33, and 34. Figure 10 is an image of step 80 through step 350. [Explanation of main component symbols] 10〇: Solar cell 110: Solar cell 120: Inner frame 310~360: Step 13