臺灣博碩士論文加值系統

本文製作一個小客車新式與舊式車牌的辨識系統，架構分為車牌定位、字元分割及字元辨識三部分。
車牌定位包含邊緣偵測、二值化、形態學、車牌候選區切割，先以Sobel垂直邊緣偵測強化影像中的邊緣，並配合中值濾波濾除雜訊，經過形態學處理後，再以輪廓追蹤演算法搜尋影像中所有輪廓，經過寬高比及高度篩選，取得車牌候選區。
字元分割包含對車牌候選區內的輪廓進行篩選，以法定寬高比初步篩除大部分非字元輪廓，計算這些輪廓的高度值眾數及輪廓起始Y座標之中位數，以計算結果，更加精確地找出所有車牌字元輪廓，最後將車牌字元以卷積神經網路進行特徵比對完成字元辨識。

This thesis designs a vehicle license plate recognition system for new and old format characters. The architecture is divided into three parts including license plate localization, character segmentation and character recognition. License plate localization includes edge detection, binarization, morphology, and possible license plate localization search. The Sobel filter is used for edge detection. It creates an image emphasizing edges. The median filter is used to remove noise from the image. After morphological processing, the contour tracking algorithm searches for all contours in the image, and obtains the possible candidates of license plates after aspect ratio and height screening. Character segmentation includes screening the contours in the candidate license plate location, preliminarily excluding most of the non-character contours by the official aspect ratio, and calculating the height mode of these contours and the median starting Y coordinate of the contour to calculate. As a result, the outline of all license plate characters can be found more accurately. Finally, the license plate characters are recognized by convolutional neural network.

摘要......................................i
Abstract.................................ii
誌謝....................................iii
目錄.....................................iv
表目錄...................................vi
圖目錄..................................vii
第一章 緒論..............................1
1.1研究動機...............................1
1.2研究方法...............................1
1.3文獻探討...............................1
1.4論文架構...............................2
第二章 研究方法 .........................4
2.1影像灰階化及直方圖均衡化................4
2.1.1灰階化..............................4
2.1.2直方圖均衡化.........................4
2.2 車牌定位..............................5
2.2.1傅立葉轉換及帶通濾波..................6
2.2.2 Sobel邊緣偵測.......................9
2.2.3 中值濾波...........................11
2.2.4 大津二值化演算法....................12
2.2.5形態學..............................13
2.3字元分割..............................16
2.4 CNN車牌字元辨識模型...................19
第三章 實驗結果與討論....................22
3.1實驗結果...............................22
3.2辨識失敗的情形..........................26
第四章 結論................................28
參考文獻...................................29
Extended Abstract.........................30
Abstract..................................30

[1]黃冠穎，“自動化車牌辨識系統設計”，國立中央大學電機工程研究所，碩士論文，2005。
[2]陳克智，“照相手機的車牌偵測與辨識”，國立中央大學資訊工程研究所，碩士論文，2011。
[3]葉本源，“適用於台灣各種車輛知車牌辨識系統”，中原大學電子工程學系，碩士論文，2006。
[4]曾士豪，“自動化辨識系統”，國立交通大學電控工程研究所，碩士論文，2016。
[5]郭傳誥，“車牌辨識系統之FPGA硬體實現”，國立中興大學電機工程學系，碩士論文，2020。
[6] Gonzalez & Woods(原著)，繆紹綱編譯，數位影像處理 2nd ed.，普林斯頓國際有限公司，2005。
[7] Nobuyuki Otsu, “A threshold selection method from gray-level histograms”, IEEE Trans. Sys. Man. Cyber, 9(1): pp.62-66, 1979.
[8] Satoshi Suzuki & KeiichiAbe, “Topological structural analysis of digitized binary images by border following”, Computer Vision Graphics & Image Processing, 30(1): pp.32-46, 1985.
[9] Y. Lecun, L. Bottou, Y. Bengio and P. Haffner, “Gradient-based learning applied to document recognition”, in Proceedings of the IEEE, 86(11): pp.2278-2324, 1998.