臺灣博碩士論文加值系統

物件偵測（Object Detection）在生活中具有諸多種應用，其中之一便是針對重複物件的計數。在現有的物件偵測演算法中，往往是針對特定目標對象，由人工設計或電腦學習相關特徵，以達到在影像中偵測特定目標的目的，但此種方法通常只能適用於特定類型的目標，並不能通用化到偵測其他種物件。
本研究延續近年來被提出，基於輪廓特徵抽取與分群的樣式偵測演算法，並在其架構上進行改良，引入具有完整物件概念的邊緣偵測方法、增加輪廓篩選機制、減少運算複雜度，使原有演算法的效能與速度都大幅提升，並達到通用化的樣式與物件偵測能力。
本研究分別針對物件偵測與物件分群兩個目標進行成果評估。物件偵測實驗的結果顯示，使用此改良演算法在原物件偵測資料集的F度量值（F-measure）可達86.48%，與改良前演算法相比提升22.37%；在本研究提供的物件偵測資料集上的F度量值也有84.32%，較改良前提升27.73%。而在物件分群的實驗中，本研究的方法在F度量值與準確度上分別達到91.43%與93.91%，較改良前的方法各自提高6.18%與14.65%。
在相同的硬體設備條件下，原方法在兩物件偵測資料集中，平均每張影像的計算時間（Second Per Frame, SPF）為130.9秒，速度較人工計算慢，本研究改良後則加速至每張平均7.2秒，使演算法更具有物件計數的實用價值。

Repeat object counting is one of the applications of object detection. Most object detection algorithms include human-designed or machine-learned feature extractor, but feature extractor designed in this way can only detect labeled targets and is hard to generalize to detect unseen objects.
Our work builds on previous proposed pattern mining algorithm, which bases on contour feature extraction and clustering to achieve general pattern detection. We focus on improving the performance and the speed of the existing algorithm and generalizing its targets to both patterns and objects by introducing noise-resistant edge detection methods, employing more contour selection mechanisms and reducing its computational complexity.
Experimental results show that in repeat object detection task, the macro precision, recall and F-measure of the algorithm improves from 80.08% to 86.54%, 53.46% to 86.43% and 64.11% to 86.48% in previous dataset, from 66.77% to 84.89%, 50.37% to 83.76% and 56.59% to 84.32% in new dataset. In object clustering task, our algorithm achieves 91.43% in F-measure and 93.91% in overall accuracy, with 6.18% and 14.65% improvement, respectively.
Under same hardware configuration, the original algorithm takes 130.9 seconds per frame in average in object detection task, which is slower than manual counting, while the improved algorithm takes only 7.2 seconds per frame, making the algorithm more practical to real object counting task.

口試委員審定書 i
誌謝 ii
摘要 iii
ABSTRACT iv
目錄 v
圖目錄 vii
表目錄 x
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機 2
1.3 研究貢獻 2
1.4 論文架構 3
第二章 相關文獻及方法探討 5
2.1 目標偵測與探勘 5
2.2 邊緣偵測相關演算法 8
2.3 數學形態學於圖像處理 13
2.4 分群演算法 15
第三章 問題定義及研究方法 19
3.1 問題定義及系統架構 19
3.2 邊緣偵測與結果處理 21
3.3 輪廓偵測與篩選 25
3.4 輪廓合併與特徵抽取 33
3.5 第一次輪廓分群與篩選 35
3.6 第二次輪廓分群與篩選 43
第四章 實驗結果與討論 48
4.1 實驗評估方式與參數說明 48
4.2 重複物件偵測實驗 49
4.3 重複物件分群實驗 57
第五章 結論與未來展望 60
5.1 結論 60
5.2 未來展望 61
參考文獻 62
附錄A 64

[1]Q. Geissmann, "OpenCFU, a new free and open-source software to count cell colonies and other circular objects," PloS one, vol. 8, no. 2, p. e54072, 2013.
[2]陳雲濤, "Automatically Count and Classify Multiple Colonies on Mobile Device Captured Images," 2016.
[3] R. Girshick, J. Donahue, T. Darrell, and J. Malik, "Rich feature hierarchies for accurate object detection and semantic segmentation," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2014, pp. 580-587.
[4] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, "You only look once: Unified, real-time object detection," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 779-788.
[5]Y. Li, L. Liu, C. Shen, and A. Van Den Hengel, "Mining mid-level visual patterns with deep CNN activations," International Journal of Computer Vision, vol. 121, no. 3, pp. 344-364, 2017.
[6]R. Zhang et al., "Mining Objects: Fully Unsupervised Object Discovery and Localization From a Single Image," arXiv preprint arXiv:1902.09968, 2019.
[7] I. Huberman and R. Fattal, "Detecting Repeating Objects using Patch Correlation Analysis," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 2903-2911.
[8]林宣佑, "Image Pattern Mining and Counting," 2017.
[9] S. Xie and Z. Tu, "Holistically-nested edge detection," in Proceedings of the IEEE international conference on computer vision, 2015, pp. 1395-1403.
[10] Y. Liu, M.-M. Cheng, X. Hu, K. Wang, and X. Bai, "Richer convolutional features for edge detection," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 3000-3009.
[11]Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient-based learning applied to document recognition," Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, 1998.
[12]K. Simonyan and A. Zisserman, "Very deep convolutional networks for large-scale image recognition," arXiv preprint arXiv:1409.1556, 2014.
[13] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778.
[14] W. Liu et al., "Ssd: Single shot multibox detector," in European conference on computer vision, 2016: Springer, pp. 21-37.
[15] J.-J. Liu, Q. Hou, M.-M. Cheng, J. Feng, and J. Jiang, "A simple pooling-based design for real-time salient object detection," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2019, pp. 3917-3926.
[16] L. Lettry, M. Perdoch, K. Vanhoey, and L. Van Gool, "Repeated pattern detection using CNN activations," in 2017 IEEE Winter Conference on Applications of Computer Vision (WACV), 2017: IEEE, pp. 47-55.
[17] T. Leung and J. Malik, "Detecting, localizing and grouping repeated scene elements from an image," in European Conference on Computer Vision, 1996: Springer, pp. 546-555.
[18]J. Canny, "A computational approach to edge detection," IEEE Transactions on pattern analysis and machine intelligence, no. 6, pp. 679-698, 1986.
[19] P. Dollár and C. L. Zitnick, "Structured forests for fast edge detection," in Proceedings of the IEEE international conference on computer vision, 2013, pp. 1841-1848.
[20]"OpenCV Tutorial, Structured forests for fast edge detection." [Online]. Available: https://docs.opencv.org/3.4/d0/da5/tutorial_ximgproc_prediction.html.
[21]"VGG16 – Convolutional Network for Classification and Detection." [Online]. Available: https://neurohive.io/en/popular-networks/vgg16/.
[22]R. M. Haralick and L. G. Shapiro, Computer and robot vision. Addison-wesley Reading, 1992.
[23]"Morphological Transformations." [Online]. Available: https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_imgproc/py_morphological_ops/py_morphological_ops.html.
[24]S. C. Johnson, "Hierarchical clustering schemes," Psychometrika, vol. 32, no. 3, pp. 241-254, 1967.
[25]K. H. Zou, A. J. O’Malley, and L. Mauri, "Receiver-operating characteristic analysis for evaluating diagnostic tests and predictive models," Circulation, vol. 115, no. 5, pp. 654-657, 2007.
[26] R. Agrawal, T. Imieliński, and A. Swami, "Mining association rules between sets of items in large databases," in Proceedings of the 1993 ACM SIGMOD international conference on Management of data, 1993, pp. 207-216.