本論文針對3D物件提出一個新的表示方法，以360度的全景影像紀錄一個3D物件，並開發以該全景影像為基礎的物件計數演算法，演算法中採用了SIFT對該影像特徵點進行匹配，接著搭配DBSCAN聚類演算法作特徵點分群。在進行特徵點匹配時，全景影像提供3D物件全面的資訊，因此在物件隨意放置(包括遮擋)的目標圖中，能夠找出與該影像相同的物件，此外還可以偵測到物件擺放的角度。論文中進行了初步實驗，驗證所提出的物件計數演算法。然而，在實驗結果中顯示角度估測的不穩定性，其精確度主要受到特徵點匹配錯誤的影響。在未來可以針對匹配的部分進行改良，以得到更準確的物件擺放角度估測。

This paper proposed a new 3D object representation using a 360-degree panoramic image and an object counting algorithm based on SIFT-based matching accompanied with DBSCAN clustering. The proposed panoramic representation records the texture information of the object from all directions. The main advantage of using this object representation is that it provides complete information of the 3D object for the matching task, so that objects can be arbitrarily placed, including occlusions, in the target image. Moreover, it is possible to derive the orientation information of the detected object. Preliminary experiments were conducted to verify the proposed object counting algorithm. However, the experimental results showed that the orientation estimation was not always stable, and the accuracy was mainly affected by the mismatched feature points. This could lead to the future work. This could be a future work to make the matching part more robust.

摘要 I
Abstract II
誌謝 III
目錄 IV
圖目錄 VI
第一章 緒論 1
1.1 研究動機與目標 1
1.2 文獻探討 2
1.3 特徵點偵測 — SIFT介紹 3
1.4 分群介紹 — DBSCAN 6
1.5 論文架構說明 7
第二章 全景圖製作方法 8
2.1 錄製影片 8
2.2 拼接影像 9
2.3 影像裁切 11
第三章 以環物影像為基礎之物件計數 13
3.1 主要流程與架構 13
3.2 特徵點偵測與特徵點匹配 15
3.3 特徵點篩選 17
3.4 特徵點分群 20
3.5 群集傾斜角度估測 26
3.6 影像相似度計算 28
3.7 群集彙整 31
第四章 實驗結果與探討 33
4.1 實驗環境與設計 33
4.2 測試基於顏色刪減特徵點的必要性 34
4.3 測試使用K-Means分群 37
4.4 以環場影像為基礎之物件計數實驗 41
第五章 結論 51
參考文獻 52

[1] H. Bay, A. Ess, T. Tuytelaars, L. Van Gool, "SURF: Speeded Up Robust Features," Computer Vision and Image Understanding (CVIU), Volume 110, Issue 3, pp. 346-359, 2008.

[2] Y. B. Brahme and P. S. Kulkarni, "An Implementation of Moving Object Detection, Tracking and Counting Objects for Traffic Surveillance System," In Pro. International Conference on Computational Intelligence and Communication Networks, Gwalior, pp. 143-148, 2011.

[3] G. Carmichael, "Matching Spherical Panoramas and Planar Photographs," Ph.D. thesis, Carleton University Ottawa (2009)

[4] P. Chattopadhyay, R. Vedantam, R. R. Selvaraju, D. Batra and D. Parikh, "Counting Everyday Objects in Everyday Scenes," In Pro. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, pp. 4428-4437, 2017.

[5] T. Y. Chuang and N. H. Perng, "Rectified Feature Matching For Spherical Panoramic Images, " Photogrammetric Engineering and Remote Sensing, Volume 84, Issue 1, pp. 25-32, January 2018.

[6] C. Del-Blanco, F. Jaureguizar and N. Garcia, "An efficient multiple object detection and tracking framework for automatic counting and video surveillance applications," IEEE Transactions on Consumer Electronics, Volume 58, Issue 3, pp. 857-862, August 2012.

[7] G. Farooque, A. B. Sargano, I. Shafi and W. Ali, "Coin Recognition with Reduced Feature Set SIFT Algorithm Using Neural Network," In Pro. FIT '16, Islamabad, pp. 93-98, 2016.

[8] L. Juan, O. Gwun, "A comparison of SIFT, PCA-SIFT and SURF," International Journal of Image Processing (IJIP), Volume 3, Issue 4, pp. 143-152, 2009.

[9] H. K. Kanagala and V. V. Jaya Rama Krishnaiah, "A comparative study of K-Means, DBSCAN and OPTICS," In Pro. ICCCI '16, Coimbatore, pp. 1-6, 2016.

[10] Y. Ke and R. Sukthankar, "PCA-SIFT: a more distinctive representation for local image descriptors," In Pro. IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), Washington, DC, USA, pp. II-II, 2004.

[11] P. S. Khude and S. S. Pawar, "Object detection, tracking and counting using enhanced BMA on static background videos," In Pro. IEEE International Conference on Computational Intelligence and Computing Research, Enathi, pp. 1-4, 2013.

[12] A. A. Khule, M. S. Nagmode and R. D. Komati, "Automated object counting for visual inspection applications," In Pro. ICIP '15, Pune, pp. 801-806, 2015.

[13] C. Li and L. Ma, "A new framework for feature descriptor based on SIFT," Pattern Recognition Letters, Volume 30, Issue 5, pp. 544-557, April 2009.

[14] D. G. Lowe, "Distinctive image features from scale-invariant keypoints," IJCV, Volume 60, Issue 2, pp. 91-110, November 2004.

[15] E. N. Mortensen, Hongli Deng and L. Shapiro, "A SIFT descriptor with global context," In Pro. CVPR '05, San Diego, CA, USA, pp. 184-190, 2005.

[16] M. S. Rahman and M. R. Islam, "Counting objects in an image by marker controlled watershed segmentation and thresholding," In Pro. IEEE International Advance Computing Conference (IACC), Ghaziabad, pp. 1251-1256, 2013.

[17] J. Sandhan, A. Mitra and V. Subramanian, "Object counting in a single surveillance image," In Pro. National Conference on Communications (NCC), Chennai, pp. 1-4, 2017.

[18] E. Schubert, J. Sander, M. Ester, H. P. Kriegel and X. Xu, "DBSCAN revisited revisited: Why and how you should (still) use DBSCAN," ACM TODS, Volume 42, Issue. 3, pp. 19-1-19-21, July 2017.

[19] K. Shiraishi, H. Kawano, H. Maeda, and N. Ikoma, "Enumeration of objects in a snap image with SIFT-based matching, " In Pro. International Conference on Soft Computing and Intelligent Systems and 9th International Symposium on advanced Intelligent Systems (SCIS/ISIS 2008), Nagoya, Japan, 2008, pp. 893-898.

[20] Y. Usui and K. Kondo, "3D object recognition based on confidence LUT of SIFT feature distance," In Pro. Second World Congress on Nature and Biologically Inspired Computing (NaBIC), Fukuoka, pp. 293-297, 2010.


[21] N. K. Verma, T. Sharma, S. D. Rajurkar and A. Salour, "Object identification for inventory management using convolutional neural network," In Pro. IEEE Applied Imagery Pattern Recognition Workshop (AIPR), Washington, DC, pp. 1-6, 2016.

[22] N. K. Verma, T. Sharma, R. K. Sevakula and A. Salour, "Vision Based Object Counting Using Speeded Up Robust Features for Inventory Control," In Pro. International Conference on Computational Science and Computational Intelligence (CSCI), Las Vegas, NV, pp. 709-714, 2016.