應用機器視覺於低對比動態X-Ray影像強化及瑕疵偵測之研究

Low Contrast X-ray Image Enhancement and Defects Classification for Real-Time Dynamic System

廖克東(Ke-Dong Liao)

指導教授 ： 江行全

元智大學/工程學院/工業工程與管理學系/碩士(2002年)

摘要

X射線非破壞性檢測技術是以X射線的照射而將產品內部的構造投影出，以達到對產品檢測的目的。但當產品是由不同材質的元件所組成時，X射線並無法以單一的強度穿透產品內的各個元件，此時，即會產生低對比影像而造成檢測上的困難。本研究之目的為針對低對比動態X-Ray影像，發展出一套影像強化及瑕疵檢測系統，以改善現行檢測上的困境。
研究中主要包含了兩方面的主題探討：（1）藉由影像增強的方法來增加影像的對比度、消除雜訊以突顯出欲檢測區域的影像；（2）瑕疵偵測，首先以相關係數為基礎的比對方法，來實現無檢測定位點的動態影像目標物及時搜尋定位，在目標物影像辨識上，使用了影像的相關性、幾何特徵及區域灰階作為影像特徵值，並以馬氏距離作為區別分析的分類器，以達到產品分類的目的。在研究結果中，本研究所提出的修正後對比伸張強化法的強化效果明顯優於各方法，相較於原始影像，此法在影像對比值提升了約40％，背景雜訊則是降低了約35％。在瑕疵偵測方面，經由兩階段的相關係數搜尋法，可即時搜尋出動態影像中目標物的正確位置。此外，本研究所提出的瑕疵檢測流程，產品分類準確率可達96％，雖然有少數的瑕疵種類可能會被誤判，但誤警率及遺漏率均不會發生，對整體檢測結果而言，不會產生危害之情形。因此，對於低對比動態X射線影像，可利用本研究所提出的方法來達到即時影像強化及瑕疵偵測。

關鍵字

X-Ray ； 影像強化 ； 樣板比對 ； 瑕疵偵測

並列摘要

X-ray inspection has wide applications in industry, because it can observe the structure inside of an object. But if a product consists of variety of materials, it would be more uncontrollable of image quality in using X-ray. Moreover, low contrast objects and noise, encounter difficulties to detect objects in X-Ray image. The focus of this research is to propose a new method for low contrast object detection based on image enhancement and defects detection algorithms.
This research can be divided into two parts: one is real time X-ray image enhancement; the other is on-line defects detection. The image enhancement included to reduce noise and to enhance some useful objects features. In defects detection: first, using image pixel correlation coefficient method to real time search an object. Secondly, the object image characteristics are extracted using image pixel correlation coefficient, geometry feature and region gray level distribution methods. At last, a product is classified using line discriminant analysis.
The research results showed that using enhancement by modified contrast stretching to compare original image, which increase image contrast more than 40% and reduce image background noise by 35%. In other words, the results of defect detection using two-phase search method could real time search object position. In product classification, the results showed that the proposed classification method obtained 96% accuracy rate. At last, this research provides an image enhancement and defects detection on low contrast dynamic X-Ray image.

並列關鍵字

X-Ray ； Enhancement ； Pattern Matching ； Defect Detection

參考文獻

Bhalla, K., N. G. Durdle, A. E. Peterson, J. Raso, D. Hill, and X. Li, “Automatic feature detection and correspondence in a stereo-vision application,” IEEE International Conference on Systems Man and Cybernetics, Vancouver, Canada, Vol. 4, pp. 3537-3542, 1995.

Campisi, P., J. C. K. Yan, and D. Hatzinakos, “Signal-dependent film grain noise generation using homomorphic adaptive filtering,” IEE Proc.-Vis. Image Signal Process, Vol. 147, No. 3, pp.283-287, 2000.

Gallegos, J. M., J. R. Villalobos, G. Carrillo, and S. D. Cabrera, “Reduced-dimension and wavelet processing of SMD images for real-time inspection,” Proceedings of the IEEE Southwest Symposium on Image Analysis and Interpretation, San Antonio, TX, USA, pp. 30-36, 1996.

Jagannathon, S., “Automatic inspection of wave solder joints using neural network,” Journal of Manufacturing Systems, Vol. 16, No. 6, pp. 389-398, 1997.

Kang, S. C., and S. H. Hong, “Design and implementation of denoising filter for echocardiographic images based on wavelet method,” Conference On.Microtechnologies in Medicine and Biology, Lyon, France, pp. 80-83, 2000.