臺灣博碩士論文加值系統

隨著硬體運算的能力日趨強健，加上掃描儀器的設備越來越普及的情況下，使得圖像識別與光學字元辨識的技術已經深入我們的生活周遭，且被廣泛的運用在市場。光學字元識別，是一種讀取影像資料的技術。一般最常見的資料輸入方法，是使用者透過鍵盤將其所需資料一一鍵入電腦中，如果資料量一但龐大時，所花費時間也會隨之倍增。倘若待輸入的資料已經是採用特定字體印刷而成，或是透過電腦列印完成者，可以透過光學字元識別系統來進行資料輸入，從而縮短所需工作時間，這對於處理龐大的印刷資料之輸入有很大的幫助。有鑑於此，為了有效的提取影像中的文字，本論文採用影像處理結合光學字元識別技術來提高其精確度，進而提出兩套系統，分別為商品廣告圖之字元識別系統和檢測自然場景中的文字識別系統。商品廣告圖之字元識別系統未來將會結合商品違法字眼檢測系統，希望藉由此系統能夠減少業者與相關單位之間的糾紛，幫助他們事先避免掉廣告圖中所出現的違法字眼；檢測自然場景中的文字識別系統主要目的是希望透過此系統來協助管理人員建檔。而在實驗結果可以看出本論文所提出之研究方法可以有效的識別影像中的文字。

With the increasing computing power and the increasing popularity of scanning instrumentation, image recognition and optical character recognition have been deeply penetrated in our life and been widely used in the market.Optical character recognition is a technique for reading image data. The most common method of data input is that the user types the required information into the computer through the keyboard. If the amount of data is large, the required working hours will also lengthen. If the information to be entered is already printed in a specific font or printed through a computer, the optical character recognition system can be used for data input, thereby shortening the required working time. In view of this, in order to effectively extract the text in the image, this paper uses image processing combined with optical character recognition to improve its accuracy, and then proposes two systems. The first system is a character recognition system for product advertisement images and the second system is a character recognition system for detecting natural scenes. The character recognition system of product advertisement images will be combined with the illegal word detection system in the future. It is hoped that the system can reduce disputes between the operators and related units and help them avoid the illegal words appearing in the advertisement images in advance. The main purpose of detecting the text recognition system in the natural scene is helping the manager to file through this system. In the experimental results, the method proposed in this paper can effectively identify the text in the image.

摘要 i
Abstract ii
誌謝 iii
目錄 v
表目錄 vii
圖目錄 viii
1、 緒論 1
1.1 研究背景 1
1.2 研究動機與目的 2
1.3 論文架構 3
2、 文獻回顧 4
2.1 光學字元識別 4
2.2 影像前處理 5
2.2.1 灰階化 5
2.2.2 模糊化 6
2.2.3 邊緣偵測 7
2.2.4 二值化 9
2.2.5 侵蝕與膨脹 10
2.3 Raspberry Pi 11
3、 研究方法 13
3.1 第一套系統－商品廣告圖之字元識別系統 13
3.1.1 影像前處理技術 14
3.1.2 尋找輪廓，圈選ROI 19
3.1.3 字元識別 20
3.2 第二套系統—檢測自然場景中的文字識別系統 20
3.2.1 打開樹莓派鏡頭，檢測帶有文字的目標物體 21
3.2.2 影像前處理技術 22
3.2.3 圈選ROI，進行文字識別 24
4、 研究結果與分析 25
4.1 系統之前置測試的研究結果與分析 26
4.2 第一套系統的研究結果 31
4.3 第二套系統的研究結果 34
5、 結論與未來展望 38
6、 參考文獻 39

[1]Hazra, T. K., Singh, D. P., & Daga, N. (2017, August). Optical character recognition using KNN on custom image dataset. In 2017 8th Annual Industrial Automation and Electromechanical Engineering Conference (IEMECON) (pp. 110-114). IEEE.
[2]Minetto, R., Thome, N., Cord, M., Stolfi, J., Precioso, F., Guyomard, J., & Leite, N. J. (2011, November). Text detection and recognition in urban scenes. In 2011 IEEE International Conference on Computer Vision Workshops (ICCV Workshops) (pp. 227-234). IEEE.
[3]Berchmans, D., & Kumar, S. S. (2014, July). Optical character recognition: an overview and an insight. In 2014 International Conference on Control, Instrumentation, Communication and Computational Technologies (ICCICCT) (pp. 1361-1365). IEEE.
[4]Hamanaka, M., Yamada, K., & Tsukumo, J. (1993, October). On-line Japanese character recognition experiments by an off-line method based on normalization-cooperated feature extraction. In Proceedings of 2nd International Conference on Document Analysis and Recognition (ICDAR'93) (pp. 204-207). IEEE.
[5]Mori, S., Suen, C. Y., & Yamamoto, K. (1992). Historical review of OCR research and development. Proceedings of the IEEE, 80(7), 1029-1058.
[6]Fu, K. S. (1976). Pattern recognition and image processing. IEEE transactions on computers, 100(12), 1336-1346.
[7]Garcia, C., & Apostolidis, X. (2000). Text detection and segmentation in complex color images. In 2000 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings (Cat. No. 00CH37100) (Vol. 4, pp. 2326-2329). IEEE.
[8]Chowdhury, S. P., Dhar, S., Das, A. K., Chanda, B., & McMenemy, K. (2009, July). Robust extraction of text from camera images. In 2009 10th International Conference on Document Analysis and Recognition (pp. 1280-1284). IEEE.
[9]Li, S., & Kwok, J. T. (2004, October). Text extraction using edge detection and morphological dilation. In Proceedings of 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing, 2004. (pp. 330-333). IEEE.
[10]Canny, J. (1987). A computational approach to edge detection. In Readings in computer vision (pp. 184-203). Morgan Kaufmann.
[11]Otsu, Nobuyuki. "A threshold selection method from gray-level histograms." IEEE transactions on systems, man, and cybernetics 9.1 (1979): 62-66.
[12]王進德（2017）。Raspberry Pi入門與機器人實作應用。新北：博碩。
[13]Karatzas, D., Gomez-Bigorda, L., Nicolaou, A., Ghosh, S., Bagdanov, A., Iwamura, M., ... & Shafait, F. (2015, August). ICDAR 2015 competition on robust reading. In 2015 13th International Conference on Document Analysis and Recognition (ICDAR) (pp. 1156-1160). IEEE.
[14]Zhong, Z., Sun, L., & Huo, Q. (2017, November). Improved localization accuracy by locnet for faster r-cnn based text detection. In 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR) (Vol. 1, pp. 923-928). IEEE.
[15]Tian, S., Pan, Y., Huang, C., Lu, S., Yu, K., & Lim Tan, C. (2015). Text flow: A unified text detection system in natural scene images. In Proceedings of the IEEE international conference on computer vision (pp. 4651-4659).

[16]Yin, X. C., Yin, X., Huang, K., & Hao, H. W. (2013). Robust text detection in natural scene images. IEEE transactions on pattern analysis and machine intelligence, 36(5), 970-983.