影像對位與直線偵測整合系統

摘要本論文係本校與全研科技股份公司的產學研究計畫，研究主要是以XXY對位平台的連續影像，分別完成快速直線偵測、特定標記對位與穩定性分析以及不同標記對位與穩定性分析等需求，以提高影像伺服定位系統對位的即時性以及穩定度。實驗裡先將輸入影像的RGB格式轉成YCbCr格式，並取出Y的亮度值，來進行灰階影像處理。直線偵測部份是以感興趣區域ROI (Region of Interest) 局部影像分割處理的概念，不同以往Hough transform必須以全域處理的複雜情形。首先由使用者角度選出(ROI)，再將(ROI)區域進行角度偵測並調整影像，接著將影像以Otsu為門檻值二值化，再分段選出每線段的平均代表位置，分段處理時，使用橫向剖面技術，統計黑色白色交替次數策略，將轉換次數過多的橫向剖面的值剃除，不僅可以避開奇異點的雜訊，卻仍保有的魯棒性，最後採用最小平方估計法(Ordinary Least Square Estimation, OLSE)[17]偵測出最接近的邊緣的直線方程式。特定對位標記(Fiducial Marker)的偵測部份，本文策略是以影像金字塔結構(Image Pyramid)概念，借由縮減取樣金字塔(Subsampling Pyramid)，將影像壓縮降低解析度和影像大小，經過縮減取樣2次後，在縮小為原圖1/16的影像上，不僅可以達到平滑化及降噪的效果，再以Otsu方法二值化後的低解析度影像，以形態學方法萃取出物件影像的邊緣特徵，再將影像逐層還原大小後，並重新調整對位，最後以絕對差異值之總和(Sum of absolute differences, SAD)比對法做比對，這樣保有工業上即時處理速度優點。實驗的結果，我們提出的方法可以降低偵測的複雜度，並達到即時的需求。直線偵測實驗，將分段由20段增加至30段，速度僅增加0.005(s)，擁有工業處理上速度的優點。

關鍵字

直線偵測；霍夫轉換；影像對位； XXY對位平台

並列摘要

This paper is an industry(CHIUAN-YAN TECHNOLOGY)- university(NUTC) cooperative research project. The main purpose of this paper are realizing real-time fast line detection and the fiducial marker detection .Furthermore, to perform the position accuracy and stability on XXY three axis positioning platform.In our practice test we utilizing the sequential images captured by(taken from) CCD and converts the RGB images to grayscale via Y of YCbCr. In fast line detection part of the project. The proposed method consists of the following four parts. Firstly, we could determine the region of interest (ROI)of captured image and rotate it’s angle might need .Then convert the gray image to binary image by otsu’s method. Secondly, we define the edge transition from black to white(b2w) or another(w2b) then decide the search direction. Thirdly, we divide the ROI boundary image into several blocks and generate representative point of block. In order to overcome noise problem we also calculate the average transition times of cross section in each block as the threshold and remove abnormal row. Finally, using least- squares method to fit a best line of the edge. In fiducial marker detection part of the project. The proposed method has the following four parts .Firstly, we take low-frequency area after 2-layer Haar discrete transform to processing with two advantages of reduce noise and smaller size. Secondly, we might choose a fiducial marker in system or determine the area of mark manually of the image then adapt morphological algorithm for extracting boundary feature. Thirdly, we matching pattern by using SAD block method and search from low-band area layer by layer that may be able to relocate for accuracy. Finally, we also make hypothesis testing in problems of contrast, incomplete marker and rotation. The results show that the proposed method can detect in real-time requirement.

並列關鍵字

Line Detection ； Hough Transform ； Image Registration ； XXY Alignment Platform

參考文獻

[5] R. C. Gonzalez and R. E. Woods, “Digital Image Processing,” Prentice-Hall, 2002.

[6] J. F. Canny, “A Computational Approach to Edge Detection,” IEEE Transaction on Pattern Analysis and Machine Intelligence, vol. 8(6), pp. 679–698, 1986.

[7] L. Ding and A. A. Goshtasby, “On the Canny Edge Detector,” Pattern Recognition, vol. 34(3), pp. 721–725, 2001.

[9] F. A. Pellegrino, W. Vanzella, and V. Torre, “Edge Detection Revisited,” IEEE transactions on systems, man, and cybernetics-part B: CYBERNETICS, vol. 34, pp. 1500-1518, 2004.

[10] C. H. Lin and Y. J. Syu, “Fast Segmentation of Porcelain Images Based on Texture Features,” Journal of Visual Communication and Image Representation, vol. 21, pp. 707-721, 2010.

國際替代計量

影像對位與直線偵測整合系統

未授權

主題瀏覽