臺灣博碩士論文加值系統

人工智慧的辨識技術於近年有十分顯著的進步，雖然應用廣泛：如車牌辨識、指紋辨識、人臉辨識、植物辨識…等，目前尚未見於建築物辨識的應用。旅人於熱門景點拍照時，常需花費很多的時間和體力去尋找一個滿意的拍攝角度。基於此需求，我們針對有特定建築物的場景，提出一即時自動辨識建築物並顯示出人像最佳拍攝角度的構想。系統的建置過程需分為三階段。第一階段：特定建築物的辨識，第二階段：建築物資料庫的建立，第三階段：人與建築物的最佳視角合成畫面。本研究目的著重在第一階段的建置，利用人工智慧（artificial intelligence AI）技術，進行建築物的自動辨識。辨識過程使用 OpenCV 提供的自行訓練機制建立建築物的 Haar-based 特徵分類器模型。
以「高雄市鳳山長老教會」為例，訓練並建立該物件的 Haar-based特徵分類器模型，再使用自行建立的 Haar-based 特徵分類器模型來偵測目標建築。本實驗測試圖片中，皆成功辨識出輸入之建築物的特徵值。利用Haar-like特徵辨識於建築成功率極高，因此Haar-like結合Adaboost 分類學習演算法不只能應用於車牌、人臉或物品，建築也是可行的。

Artificial intelligence recognition technology has made remarkable progress in recent years. Although it is widely used, such as license plate recognition, fingerprint recognition, face recognition, plant recognition, etc., it has not yet been used in building recognition.
While travelers take pictures at tourist attraction, they often spend a lot of time and energy to find a satisfactory shooting angle. Based on this requirement, we propose a concept of real-time automatic identification of buildings and display of the best angle of shot for portraits in scenes with specific buildings. We divided construction of the system into three stages. The first stage: identification of a specific building; the second stage: the establishment of the building database; the third stage: the best perspective synthesis of people and architecture. The purpose of this paper is to focus on the construction of the first stage, using artificial intelligence technology to identify architecture automatically. The identification used the self-training Haar-based feature mechanism by Open CV.
Take " Presbyterian Church of Taiwan Fengshan Church " as an example, train and build the Haar-based feature classifier model of the object, and then use it to detect the target building. In the test images of this experiment, the targets of the input buildings were successfully recognized. Using Haar-like features to identify architecture has a very high success rate, so Haar-based feature can not only be applied to license plates, faces or objects, but also buildings.

Chapter1 緒論
1.1研究背景與動機
1.2研究目的
1.3論文架構

Chapter 2 文獻探討
2.1特徵擷取(Haar-like feature Extraction)
2.2 特徵訓練(積分圖Integral Image的計算)
2.3 Adaboost 分類學習演算法
2.4 Cascade Classifier階層式分類器

Chapter 3 實作方法與架構
3.1 選擇愈辨識建築之特徵
3.2 影像處理
3.3 分類器訓練過程

Chapter 4 實驗結果分析與討論
4.1環境設置
4.2實驗結果
4.3 辨識系統測試
4.4 實驗結果討論

Chapter 5 結論與未來研究

參考文獻

參考文獻
[1] C.P. Papageorgiou ,M. Oren and T. Poggio, A general framework for object detection Sixth International Conference on Computer Vision,1998.
[2] P. Viola and M. Jones, “Rapid object detection using a boosted cascade of simple features,” IEEE Conference on Computer Vision and Pattern Recognition, PP. 511-518,2001
[3] Y.Frecund and R.E Schapire, A Short Introduction to Boosting. Journal of Japanese Society for Artificial Intelligence, Volume 5, NO 5 PP. 771-780,1999.
[4] R. Lienhart and J. Maydt , 2002. An extended set of Haar-like features for rapid object detection Proceedings. International Conference on Image Processing. PP. 511-518
[5] Y.Frecund and R.E Schapire, A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting. Journal of Computer and System Sciences. Volume 55, Issue 1. PP. 119-139,1996.
[6] 黃瀚文、張國清、李傳仁。2015. 應用半監督式參數更新之車輛偵測系統。 The 20th National Conference on Vehicle Engineering, Nov. 13, 2015, Da-Yeh University., Changhua, Taiwan, R.O.C. B – 010
[7] 鄧文淵。2020. Python機器學習超進化：AI影像辨識跨界應用實戰。初版，Ch3 2-26。台北：碁峰。
[8] S. Suria, A. Sankaranb, M. Vatsac and Richa Singh, Improving face recognition performance using TeCS2 dictionary. Pattern Recognition Letters. Volume 145. PP. 88-95,2021.
[9] 許富淵。2021. Application of face recognition in a home door locking system. 碩士論文。彰化：國立彰化師範大學資訊工程學系物聯網碩士班。
[10] 王馗行。2016. 植基於Haar-like特徵與AdaBoosting之車輛偵測模型。碩士論文。台中：國立勤益科技大學資訊管理系。