傳統電容辨識方法是依靠人工來檢查一張影像中有幾層電容層。但利用人眼來檢測缺乏效率且不可靠，因此自動化辨識在現代的工業檢測中已是一種趨勢且被廣泛的利用。本篇論文實現一個積層陶瓷電容影像的辨識系統。電容影像分析的困難主要包含沒有參考樣版、電容層附近雜訊的影響等。

　　在論文中，我們結合基本影像處理和影像轉換技術來找出影像中電容層的層數。我們的檢測方法包含三部分：(1)傾斜偵測以及影像校正，(2)區塊分割及(3)電容層數計算。檢測過程描述如下：
　　在第一部份利用中心點計算、方向偵測所得到的資訊隨後結合區塊搜尋法來找出電容層傾斜角度，然後將輸入影像經由旋轉校正。在第二部份中，先從校正後影像得到投影量，然後用小波轉換和分類演算法來切割電容層邊界範圍。第三部份我們再利用快速傅立葉轉換和間距偵測來找出電容層的間距，最後結果則是利用間距和電容層範圍來進行電容層數計算。

　　實驗裡，我們對8種產品共256個影像進行辯識，若辨識結果和真正的層數之間誤差的絕對值小於0.5就可視為正確，實驗結果顯示我們的方法可以讓256個影像計算出來的誤差均小於0.5。平均檢測一張電容影像所需的時間為443.4 ms，平均影像大小為300×300。

　　The traditional capacitor recognition is manual, unreliable, and inefficiency. Automatic recognition would therefore have great benefits and have been widely adopted for various automatic visual inspections in today’s industry. This thesis implements a recognition system to calculate how many layers in a MLCC image. The main difficulties of the problem includes: without template image, noise near each layer…, etc.

　　The proposed algorithms combined basic image processing and image transform techniques in our recognition system to find the number of layers. Our algorithm consists of three parts: (1) skew detection and alignment, (2) object segmentation and (3) layer number calculation. In the first part, by using the information from centroid and direction, block matching algorithm is performed to find the skew angle and then the image is aligned via a rotation matrix. In the second part, projection is performed first. After that we use wavelet frames and fuzzy c-means to locate the layer boundary in the image. At last part, fast Fourier transform and pitch detection is performed to find pitch between layers. The number of layers is computed by combing the pitch and the layer boundary.

　　In the experiments, we used eight products with total 256 images as test samples. The error is defined as the difference between the actual layer number and the number computed from the computer system. According to our experimental results, the errors are less than 0.5 for all 256 images. As for the processing time, 443.4 ms in average is required by one image and the average of the image size is 300×300.

[1]N. Otsu, “A threshold selection method from gray-level histogram,” IEEE Transactions on Systems, Man, and Cybernetics, SMC-9, pp. 62-66, 1979.

[2]R. C. Gonzale and R. E. Woods, Digital Image Processing, United States of America, 1993.

[3]J. Canny, “A computational approach to edge detection,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 8, no. 6, pp. 679-697, November 1986.

[4]V. Fotopoulos and A. N. Skodras, “Smae: an improved block matching criterion,” IEEE International Conference on Electronics, Circuits and Systems, vol. 3, pp. 519-522, 1998.

[5]J.-N. Kim and T.-S. Choi, “A fast three-step search algorithm with minimum checking point using unimodal error surface assumption,” IEEE Transactions on Computer Electronics, vol. 44, no. 3, pp. 638-648, August 1998.

[6]J. Lu and M. L. Lious, “A simple and efficient search algorithm for block-matching motion estimation,” IEEE Transactions Circuits System Video Technology, vol. 7, pp. 429-433, April 1997.

[7]R. C. Gonzales and R. Safabakhsh, “Computer vision techniques for industrial inspection and robot control: a tutorial overview,” IEEE Computer, pp. 400-420, May 1986.

[8]A. W. Paeth, “A fast algorithm for general raster rotation,” in Proc. Graphics Interface Vision Interface, pp. 77-81, May 1986.

[9]M. Unser, P. Thévenaz and L. Yaroslavsky, “Convolution-based interpolation for fast, high-quality rotation of images,” IEEE Transactions on Image Processing, vol. 4, no. 10, pp. 1549-1560, October 1995.

[10]S. Chien and Y. M. Baek, “Hierarchical block matching method for fast rotation of binary images,” IEEE Transactions on Image Processing, vol. 10, no. 3, pp. 483-489, March 2001.

[11]M. Unser, “Texture classification and segmentation using wavelet frames,” IEEE Trans. Image Process, vol. 4, no. 11, pp. 1549-1560 November 1995.

[12]T. Cheang and C.-C. J. Kuo, “Texture analysis and classification with tree-structured wavelet transform,” IEEE Trans. Image Process, vol. 2, no. 4, pp. 429-441, October 1993.

[13]Martin Vetterli, Jelena Kovačević, Wavelets and Subband Coding, Prentice-Hall, NJ, 1995.

[14]S. Mallat, “A theory of multiresolution signal decomposition: The wavelet representation,” IEEE Trans. Patt. Anal. Machine Intell., vol. 11, no. 7, pp. 674-693, 1989.

[15]S. Mallat and S. Zhong, “Characterization of signals from multiscale edges,” IEEE Trans. Patt. Anal. Machine Intell., vol. 14, no. 7, pp. 710-732, July 1992.

[16]A. Laine and J. Fan, “Frame representations for texture segmentation,” IEEE Trans. Image Process., vol. 5, no. 5, pp. 771-780, May 1996.

[17]Q. Pan, L. Zhang, G. Dai and H. Zhang, “Two denoising method by wavelet transform,” IEEE Trans. Image Process., vol. 47, no. 12, pp. 3401-3406, December 1999.

[18]I. Daubechies, Ten Lectures on Wavelet. Philadelphia: Soc. Ind. Applied Math., 1992

[19]K. L. Laws, “Rapid texture identification,” Proc. SPIE, vol. 238, pp. 376-380, 1980.

[20]J. S. Jang, C. T. Sun, E. Mizutani, Neuro-Fuzzy and Soft Computing, Prentice-Hall International, NJ, 1997.

[21]J. C. Bezdek, Pattern Recognition with Fuzzy Objective Function Algorithm, Plenum Press, New York, 1981.

[22]吳盈璋, “用快速定位演算法做晶片印碼品質檢測,” 碩士論文, 國立成功大學電機工程研究所, 1999.

[23]陳鵬仁, “基因式特徵粹取應用於使用小波理論之紋路影像切割,” 碩士論文, 國立成功大學電機工程研究所, 1998.

[24]Milan Sonka, Vaclav Hlavac, Roger Boyle, Image Processing, Analysis, and Machine Vision, United States of America, 1998.

[25]Ioannis Pitas, Digital Image Processing Algorithms, Prentice-Hall International, NJ, 1993.

[26]Alan V. Oppenheim, Ronald W. Schafer, Discrete-Time Signal Processing, Prentice-Hall International, NJ, 1999.

[27]G. T. Uber, “Illumination methods for machine vision,” Proceedings of
SPIE, 1005, pp. 93-102, 1986.