臺灣博碩士論文加值系統

由於科技的發展快速，影像處理的技術趨於成熟也越來越普遍，且業界的一些公司也導入自動檢測與辨識的相關設備，其中影像處理與辨識的應用被視為一門重要的技術科技。本研究主要是檢查單眼相機在生產組裝過程之外觀缺陷，在生產線當中會因人為的因素導致產品變形以及遺漏工程或有功能性之重要尺寸錯誤，該錯誤會造成市場上的抗議而間接造成公司的莫大損失，因此產品在組裝製程當中要把所有會造成尺寸錯誤或產品特徵缺失的部分就必須確保，而大部分可藉由製程能力來管控，如間隙、毛邊…等外觀類，必須由最終包裝出貨前做全檢的動作，避免不良品的流出。這項研究是注重於檢查相機產品上的功能鍵印刷及螺絲之缺陷，研究上使用CCD鏡頭搭配環形光來擷取影像，並使用NI LabVIEW影像計算軟體作影像二值化、影像相減、影像型態學、像素換算、Pattern Matching，其中Pattern Matching演算法包含有影像灰階調整、直方圖均衡化、高斯金字塔、像素強度、圖像匹配等方式處理。而產品尺寸量測是以良品影像與檢測平台背景圖，二者相減所得的影像，利用二值化與影像型態學所過濾出來2D平面圖像，並透過Regions of Interest (ROI)得知該產品影像2D平面圖像的長與寬像素，再經由相機校正後進行像素換算便可得知實際的尺寸，而特徵檢測方面則是使用Pattern Matching演算法，其作法是從NI LabVIEW影像計算軟體取得無瑕疵產品樣板圖像後，且在樣板圖像裡設定為匹配之產品特徵點，將特徵點圖像執行直方圖均衡化後，使特徵點的圖像灰度均分，並運用高斯金字塔之下採樣，使圖像尺寸縮減，且縮減後的圖像執行直方圖，將直方圖呈現出的像素強度與目標匹配的像素強度之間作百分比，將百分比結果相關數值獲得score，由此score做為檢測產品特徵有無瑕疵的重要判斷。

Due to the rapid development of technology, the technology of image processing has recently become more common. Some company also uses some equipments that helps to atomatically dectect and recognize the defects on the products. This study foucus on inspecting the defects on single lens reflex (SLR) cameras. In the processing of assembling SLR cameras, some personal factors may lead to dirstortion, lacks or, which will lead to great loss on revenus. In the process of assembling SLR cameras , inspectors find sthe factors that will lead to defects like gaps or trimming. Then, do the overall inspection before packaging the products. This survey focuses on the printing on the function keys and use the visual detection to find out the defects on the screws. Use the CCD lens and ring light to capture image. Then, use NI LabVIEW to deal with difference images, binarization, image morphology, and pixel Conversion. The algorithm pattern matching includes gray scaling adjustment, histogram equalization, Gaussian pyramid, pixel intensity and image matching. Using the images of products to XOR the background images to binarize the images. Then, the autual pixels in regard to the lengths and widths of the produts through regions of interest (ROI). And use pixel coversions to get the correct size of the cameras. In terms of Image feature detection, using Pattern Matching to detect the characteristics of the products. Get the images of no defected SLR products from NI LabVIEW as a template.Using histogram equalization algorithm to equally gray the template. Then, use Gaussian pyramid to reduce the size of the template. After that, implement the template to histogram. Compare the pixel intensity of template with the the pixel intensity of the target object, it will forms a percentage cotrast. Then, use image matching to form a score. The score is the reference to identify the product.

摘要.......................................i
Abstract...................................ii
誌謝 ...................................iii
目錄.......................................iv
表目錄......................................vi
圖目錄......................................vii
第 一 章 緒論................................1
1.1 研究背景與動機............................1
1.2 研究目的..................................2
1.3 論文架構..................................2
第 二 章 文獻回顧..............................3
2.1 影像尺寸量測相關研究.........................3
2.2 影像視覺辨識相關研究.........................5
第 三 章 研究架構與方法..........................8
3.1 研究架構與流程...............................8
3.2 硬體及系統架構...............................8
3.2.1 取像硬體設備..............................9
3.2.2 環形光源硬體設備...........................9
3.3 影像視覺尺寸量測系統建置......................10
3.3.1 取像系統之光源設計與架設....................10
3.4 相機失真校正與補償...........................12
3.5 影像處理演算法...............................14
3.5.1 灰階轉換與影像去背......................... 14
3.5.2 二值化....................................15
3.5.3 影像型態學................................16
3.6 關鍵區域(Regions of Interest, ROI)..........17
3.7 二維影像視覺幾何特徵.........................18
3.7.1 影像灰度調整..............................19
3.7.2 直方圖均衡化(Histogram Equalization, HE)..19
3.7.3 高斯金字塔(Gaussian Pyramid)..............21
3.7.4 圖像匹配..................................23
3.8 產品尺寸量測................................24
3.9 產品瑕疵檢測................................25
第 四 章 實驗結果...............................26
4.1 標準件塊規量測..............................26
4.1.1 標準件重複精度............................27
4.2 影像軟體處理................................28
4.2.1 彩色轉灰階與影像去背.......................28
4.2.2 影像二值化與影像型態學.....................29
4.3 幾何尺寸驗證................................29
4.3.1 產品尺寸量測..............................29
4.3.2 尺寸比較差異..............................30
4.4 影像比對瑕疵................................30
4.4.1 影像灰階調整與直方圖均衡化.................30
4.4.2 高斯金字塔................................32
4.4.3 高斯模糊與像素強度.........................33
4.5 產品實體瑕疵檢測.............................33
4.5.1 樣品影像取樣特徵...........................33
4.5.2 影像比對瑕疵判斷...........................34
4.5.3 檢測瑕疵品結果.............................35
第 五 章 結論與未來展望...........................37
參考文獻.........................................38
Extended Abstract...............................40

[1]Laigang Zhang,Qili Yang,Qun Sun,Deying Feng,Ying Zhao,”Research on the size of mechanical parts based on image recognition”, Journal of Visual Communication and Image Representation Volume 59, February 2019, Pages 425-432
[2]Yu-cun Zhang,Jun-xia Han,Xian-bin Fu,Fu-li Zhang,” Measurement and control technology of the size for large hot forgings”, Volume 49, March 2014, Pages 52-59
[3]Gangfeng Xiao,Yongting Li,Qinxiang Xia,Xiuquan Cheng,Weiping Chena,” Research on the on-line dimensional accuracy measurement method of conical spun workpieces based on machine vision technology”, Volume 148, December 2019, 106881
[4]Zhu Rong,Zhu Li,Li Dong-nanc,” Study of color heritage image enhancement algorithms based on histogram equalization”, Volume 126, Issue 24, December 2015, Pages 5665-5667
[5]Joanna Isabelle Olszewska,” Active contour based optical character recognition for automated scene understanding”, Neurocomputing ,Volume 161, 5 August 2015, Pages 65-715 August 2015, Pages 65-71
[6]M. Ren, J. Yang, H. Sun,” Tracing boundary contours in a binary image”, Image Vis. Comput., 20 (February (2)) (2002), pp. 125-131
[7]R. Brunelli,” Template Matching Techniques in Computer Vision: Theory and Practice”, John Wiley and Sons (2009)
[8]Yufeng Shu,YongGang Chen,Changwei Xiong,” Application of image recognition technology based on embedded technology in environmental pollution detection”, Microprocessors and Microsystems,Volume 75, June 2020, 103061
[9]Mohammadamin Kashiha, Claudia Bahr, Sanne Ott,Christel P.H.Moonsb,Theo A.Niewold, F.O.Ödbergb, Daniel Berckmansa, ” Automatic identification of marked pigs in a pen using image pattern recognition”,Computers and Electronics in AgricultureVolume 93, April 2013, Pages 111-120
[10]圖片網址http://140.112.14.7/~measlab/course/2013/量測技術應用/1-光學影像之基礎%20(NTU).pdf
[11]圖片網址http://nova.bime.ntu.edu.tw/~ttlin/Course01/lecture_notes/C1_LECTURE_NOTE_06(2%20in%201).pdf
[12]圖片網址http://gh-lin.blogspot.com/2018/02/morphology.html
[13]圖片網址https://www.cs.ccu.edu.tw/~damon/photo/proj/2013/601415056/601415056.pptx
[14]圖片網址 謝文達, 2011 ”實用得影像對比增強效能評估模式”逢甲大學 ，資訊電機工程，碩士學位論文。
[15]圖片來源David G. Lowe. “Distinctive Image Features from Scale-Invariant Keypoints” International Journal of Computer Vision volume 60, pages91–110(2004)
[16]蔡治平, 2013 “影像視覺之光譜照度分析與物件特徵辨識”, 國立虎尾科技大學，機械與電腦輔助工程系，碩士學位論文。
[17]楊承翰, 2014“基於FPGA之物件影像自動化辨識與檢測系統”, 國立虎尾科技大學，機械與電腦輔助工程系，碩士學位論文。
[18]許嘉洧, 2018“基於格雷碼結構光輔助工具機之工件自動定位”, 國立虎尾科技大學，機械與電腦輔助工程系，碩士學位論文。
[19]簡駿宥，2019，”基於物件姿態估測導引機械手臂智慧取放與加工系統”，國立虎尾科技大學，機械與電腦輔助工程系，碩士學位論文。