透過您的圖書館登入
IP:18.220.59.69
  • 學位論文

曲型車用後視鏡之自動化成像歪斜瑕疵檢測

Automated Surface Distortion Defect Inspection of Curved Car Mirrors

指導教授 : 林宏達
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


曲型鏡面相較於平面鏡面具有較高反射性、增加鏡面成像視野等特性,目前已廣泛應用於汽機車用後視鏡及道路週邊之安全反射鏡,讓駕駛人員透過鏡面反射更能掌握行車路況而降低行車的危險性。在曲型車用後視鏡的生產製程中,後視鏡在製品常因烘曲時受到爐內溫度變化及玻璃成形拉曲等因素影響而產生歪斜瑕疵。具歪斜瑕疵的鏡面其對物件的成像會產生明顯的變形,而變形所造成的歪斜程度不易以量化方式表示,加上曲型鏡面具有高反射之特性,對於不同程度之歪斜瑕疵的認定,更增加分辨上的難度。因此本研究提出一曲型鏡面玻璃的自動化歪斜瑕疵檢驗方法,利用曲型鏡面具高反射之特性,使用棋盤式方格狀標準檢驗板反射於鏡中之成像進行影像擷取,其成像圖案會因受測鏡面是否具有不同程度歪斜瑕疵而產生相對應的成像變形。 本研究利用棋盤式標準檢驗板之60個線段的交點將歪斜瑕疵之變形程度進行量化,並計算各交點與中心點的距離為特徵值,接著比較此特徵值與正常成像所對應之距離,並計算其各對應的距離歪斜偏移量。最後將此歪斜偏移量透過微量偏移管制法偵測,其可有效判斷鏡面玻璃是否存在歪斜瑕疵。本研究實驗樣本分為兩階段實驗,第一階段以小樣本數量85張,用以決定較佳參數的範圍,另外,在第二階段為大樣本實驗,主要使用第一階段較佳參數範圍實驗,該階段使用386張待測影像進行曲型鏡面玻璃之歪斜瑕疵偵測,實驗結果顯示本研究所提方法其正常樣本誤判率為4.41%,而瑕疵檢出率達98%,此自動化方法具有顯著性的檢驗效果。

並列摘要


Comparing with plane car mirrors, curved car mirrors have characteristics of higher reflectance and wider field of view. Currently, the curved mirrors have been widely used in vehicle rearview mirrors and security mirrors on the driving roads and make drivers have better fields of views and driving information. In the production process of curved car mirrors, mirrors with surface distortion defects results from the unstable temperature changes of ovens and inappropriate control of over-flow fusion process. It is not easy to measure the magnitudes of distortion defects on curved car mirrors. Furthermore, the curved mirrors with the property of higher reflection increase the difficulty of discrimination of surface distortion defects on car mirrors. This study proposes a novel approach based on small-shift control scheme to inspect surface distortion defects on curved car mirrors. In order to quantize the deformation (degree of distortion) of a car mirror, a standard inspection pattern (checkerboard grids) is designed to reflect the pattern on a testing car mirror for image acquisition. The reflected pattern image of a defective mirror with distortion is compared with that of a normal mirror for quantifying the deformation and locating the distortion defects. We first detect the intersection points of the standard inspection pattern, then measure the distances of the intersection points from the origin, and calculate the distance deviations of the corresponding intersection points between the defective and normal images. Finally, we apply the small-shift control schemes, the cumulative sum (CUSUM) method and the exponentially weighted moving average (EWMA) method, to judge the existence of the distortion defects based on the accumulative deviation distances. Experimental results show that the proposed methods achieves a high 98% probability of correctly discriminating distortion defects on curved car mirrors.

參考文獻


[50] 柯紳彥,「應用衛星影像決定波向線及估算水深之初步探討」,碩士論文,交通大學土木工程學系研究所,張憲國教授指導,(2012年)。
[51] 黃智政,「應用區塊離散餘弦轉換搭配灰預測模式檢測車用平面鏡面玻璃可視瑕疵」,碩士論文,朝陽科技大學工業工程與管理所,林宏達教授指導,(2009年)。
[55] 陳韋吉,「透明玻璃之自動化可視瑕疵檢測」,碩士論文,朝陽科技大學工業工程與管理研究所,林宏達教授指導,(2011年)。
[48] 李英達,「TFT-LCD 小尺寸面板之玻璃表面瑕疵的自動強化及檢測」,碩士論文,成功大學資訊工程學系,連震傑教授指導,(2005年)。
[43] Yang, Ni, Wang, Yu-tian, Lin, Xiao-lin, Lv, Jiang-tao, and Pan, Zhao, “Study on a Novel Laser Measurement System for Float Glass Thickness”, Proceedings of 2008 IEEE International Conference on Mechatronics and Automation, 978-1-4244-2632-4/08, pp. 179-184, (2008).

被引用紀錄


林宥凱(2015)。車用後視鏡之輪廓瑕疵檢測〔碩士論文,朝陽科技大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0078-2502201617130116

延伸閱讀