CAD(Computer Aided Design)一直是近代工業設計上最常被使用的工具，不論是建築、電子元件、3C產品…等等，而CAD所對應的檔案格式有dxf、dwg、cad…，這些檔案都是由向量格式的資料結構所構成，此外還有近幾年非常熱門並且在人類生活中占據了很重要部分的一項網路服務─地圖，例如最知名的Google Maps的部分功能所使用的地圖圖資，都是由衛星影像取得柵格數據再向量化成一般民眾從電腦螢幕上看到的影像，基於越來越多科學上的應用，影像向量化所需處理資料量也跟著與日增長，除了輸出結果的正確性外，效率亦是被考量的重點之一。
在本研究中的流程區分為幾個階段，首先進行影像分割，再將各個分割影像進行邊緣偵測取得各影像的邊緣資料，透過邊緣資料來偵測各影像中的幾何圖形，以上的處理將會透過多工的方式來進行，藉由多工的方式來提升時間上的效率，接著將各個分割影像的幾何圖形合併回一張完整影像，考量到不為特定影像領域所限制的情境，因此我們採用一般影像格式為輸入資料，在影像邊緣偵測上，將會比較不同的邊緣偵測演算法來決定何種演算法較為適用在影像向量化上，並且說明為何最後採用Canny Operator為本次研究中的邊緣偵測演算法，Hough Transform 是本研究中所採用的幾何圖形偵測演算法，經過驗證
實驗數據中，明顯呈現出分割影像的多工運算在執行時間上，優於單一影像的單工運算，而分割運算所產生的像素誤差和幾何圖形誤差，與單一影像的結果相比較後的誤差小於一個百分比，另外我們還將所提方法在實作中的程式，驗證了物體的實際尺寸，完全可以達到工業標準。

CAD (Computer Aided Design) has been the most commonly used tool of modern industrial design, and its application covers construction, electronics components, 3C products and so on. Moreover, the corresponding CAD file formats have dxf, dwg, cad, etc. The data structure of these files are constituted by a vector format, in addition, a network services is a highly popular and occupied a strongly important part for a life of human is in recent years ─ map, for example, the most famous part of the function of Google Maps uses map data which is raster data, It is acquired by the satellite images which is vectored to general users to see from the image on the computer monitors, base more and more creative applications on science. The amount of data of image processing required for vectorization is increasing steadily, in addition to the correctness of the output results, the efficiency is also one of the key considerations.
In the process of this research is divided into several stages, first is for image segmentation, and then performs edge detection of each divided image data to get the image edge through which detects the geometry, the above process will be performed with multi-task via which the efficiency of process time is improved. The geometry of each division will be merged back into one complete image, it should be considered that is not restricted to a particular field of image process of situations, therefore the input image data of this research will take the general format, for example jpg, bmp, tiff, etc.
On the image edge detection, this research will compare different edge detection algorithm to determine which algorithm is more applicable to the vectorization in the image, and explains why the final selection is Canny Operator as the edge detection algorithm of this research. Hough Transform is used as the geometry detection algorithm of this research, after experimental verification is valid and applicable to the research.
The experimental data clearly displaying multitasking performance of a spilt image on the execution time is better than a single image of a single task operation, and the error is less than one percentage that is the process result produced the dividing pixel errors and geometry errors compared with a single task process does, the research proved that the proposed methods can effectively improve the efficiency and maintain correctness. In addition, we implemented the methods we proposed for verifying the actual size of objects, whose results can meet the industry standards.

[1]James D. Murray and William vanRyper, Encyclopedia of Graphics File
Formats, 2nd Edition. O'Reilly Media, 1996.
[2]Injection molding: http://en.wikipedia.org/wiki/Injection_molding
[3]Rafael C. Gonzalez and Richard E. Woods, Digital Image Processing (3rd
Edition). Prentice Hall, 2007.
[4]Chris Solomon and Toby Breckon, Fundamentals of Digital Image Processing: A Practical Approach with Examples in Matlab. Wiley, 2011.
[5]William K. Pratt, Introduction to Digital Image Processing. CRC Press, 2013.
[6]Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest, Introduction to
Algorithms, Second Edition. The MIT Press, 2001.
[7]Linear equation, http://en.wikipedia.org/wiki/Linear_equation
[8]Polyline, http://en.wikipedia.org/wiki/Polygonal_chain
[9]The Distance Formula, http://www.purplemath.com/modules/distform.htm
[10]Img2CAD, http://www.img2cad.com
[11]Emgu CV,http://www.emgu.com/wiki/index.php/Main_Page
[12]Mohamad Hanif Md Saad, Wilson Yong Wai Soon, Muhammad Faiz Mohd Shukr, W Mimi Diyana W Zaki, Aini Hussain, “QIMS : An Integrated Image Measurement and Management System For Engineering And Scientific Application,” 2014 IEEE 10th International Colloquium on CSPA, Cambridge, MA, March 2014 , pp. 292-296.
[13]Yongwei Zhang, Xitao Zheng, and Junsheng Zhong, “A New Morphological Ontology MethodFor Carp Based On Digital Image Measurement,”
International Conference on AICI, vol. 2, Sanya, Oct. 2010, pp. 23-27.
[14]Min Li, Wei-Wei Zhang, and Biao Ma, “The measurement system of hardware's shape and dimension based on image processing method ,“ International Conference on Machine Learning and Cybernetics,
vol. 6, Baoding, July 2009, pp. 3150-3154.
[15]Map Hu, Xiaodong, Shen, Zhanfeng, Luo, Jiancheng, and Xia Liegang, “An Adaptive Parallel Vectorization Method for RS Segmented Raster Map,” 2nd
International Conference on ICISE, Hangzhou, China, Dec. 2010, pp. 3514-3517.
[16]Chen R X, Zhao Z M, and Pan J. “A Fast Method of Vectorization for RS Classified Raster Map.” Journal of Remote Sensing, pp. 326-331, 2006.
[17]Fischer, J. and Pribula, O., “Precise Subpixel Position Measurement with Linear Interpolation of CMOS Sensor Image Data,” IEEE 6th International
Conference on IDAACS, vol. 1, Prague, Sept. 2011, pp. 500-504.
[18]Wei Zhang, Wenjuan Wang, Jingchun Wang, and Zhihua Xiong, “A new image measurement method for wheel hub,” IEEE International Conference
on IST, Beijing, Oct. 2013, pp. 75-78.
[19]Shujun Fu, Qiuqi Ruan, and Wenqia Wang, “A nonlinear PDE based edge detection for image measurement,” International Conference on ICCA '05,
vol. 2, pp. 1028-1033, 2005.
[20]Pribula, O. and Fischer, J., “Real time precise position measurement based on low-cost CMOS image sensor: DSP implementation and sub-pixel measurement precision verification,” 18th International Conference on
IWSSIP, Sarajevo, June 2011, pp. 1-4.
[21]Bi Guihong, “Relative Image Measurement on Cigarette Filter Tip,” 8th
International Conference on ICEMI ’07, Xi'an, July 2007, pp. 2961-2964.
[22]Slusanschi, E. and Lupescu, G., “mage vectorization on modern architectures,” IEEE International Conference on ICCP, Cluj-Napoca, Sept.
2013, pp. 295-298.
[23]Zhang Xiao-xian, Zheng Guo-xun, Fu Hao-hai, and Cao Li-ying, “Research on image vectorization based on mathematical morphology,” International
Conference on MEC, Jilin, Aug. 2011, pp. 1179-1182.
[24]Canny Edge Detection, http://www-scf.usc.edu/~boqinggo/Canny.htm
[25]Sobel Edge Detector, http://homepages.inf.ed.ac.uk/rbf/HIPR2/sobel.htm
[26]Hough Transform, http://opencv-python-tutroals.readthedocs.org/en/latest/py_tutorials/py_imgp
roc/py_houghlines/py_houghlines.html
[27]Hash Tables, http://www.sparknotes.com/cs/searching/hashtables/section1.html
[28]The Benefits of Automation, http://www.robots.com/articles/viewing/advantages-of-automation