臺灣博碩士論文加值系統

此篇論文探討背景消去演算法，不同於傳統的作法，我們將目標影像與背景影像丟給卷積類神經網路來做訓練。重點在於背景影像的產生，背景影像並非時序的中值濾波所產生的，而是由高斯混合模型所產生，在此基礎下，背景的可信度將有卓越的提升。而我們也探討灰階影像與彩色影像對訓練結果的影響，以及是否讓卷積類神經網路產生的前景遮罩參與高斯混合模型的背景生成。多方探討下，我們發現在選取的2014 ChangeDetection.net資料庫中，展現出良好的結果，優於當前的IUTIS-5、PAWCS、SuBSENSE等方法。

This paper aims to analyze background subtraction algorithm. Different from tradition methods, we feed the trained network with the target and background images. Focusing on how to get the background images. Not using the temporal median filter. We use the Gaussian mixture models to produce background images. In this way, the accuracy of background images increases. We also research the difference between grayscale and RGB images. And whether adding the foreground masks from the convolutional Neural Networks to the Gaussian mixture models or not. Experiments lead on 2014 ChangeDetection.net dataset show that our proposed method outperforms several state-of-the-art methods, including IUTIS-5, PAWCS, SuBSENSE and so on.

口試委員審定書 #
誌謝 i
摘要 ii
Abstract iii
目錄 iv
附圖目錄 vi
附表目錄 vii
第一章 緒論 1
1.1 前言 1
1.2 研究動機與目標 2
1.3 論文架構 3
第二章 文獻探討 4
第三章 實驗流程與架構 7
3.1 輸入資料前置處理 7
3.2 背景模組 7
3.3 訓練資料 9
3.4 類神經網路架構 10
第四章 實驗結果 12
4.1 前置作業 12
4.2 數據探討與比較 13
4.3 結果展示 15
第五章 研究方法探討與驗證 20
5.1 訓練RGB影像 20
5.2 前景遮罩參與高斯混合模組來產生背景 20
5.3 驗證一般背景 22
第六章 結論 24
參考文獻 25

[1] M. BRAHAM and M. VAN DROOGENBROECK. Deep Background Subtraction with Scene-Specific Convolutional Neural Networks. In IEEE International Conference on Systems, Signals and Image Processing (IWSSIP), Bratislava, Slovakia, pages 1-4, May 2016. doi:10.1109/IWSSIP.2016.7502717, BiBTeX entry.
[2] N. Goyette, P.-M. Jodoin, F. Porikli, J. Konrad, and P. Ishwar, “A novel video dataset for change detection benchmarking,” IEEE Trans. Image Process., vol. 23, pp. 4663–4679, Nov. 2014.
[3] Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, “Gradient-based learning applied to document recognition,” Proc. of IEEE, vol. 86, pp. 2278–2324, Nov. 1998.
[4] P. Gil-Jimenez, S. Maldonado-Bascon, R. Gil-Pita, H. Gomez-Moreno, “Background pixel classification for motion detection in video image sequences“, International Work Conference on Artificial and Natural Neural Network, IWANN 2003, Volume 2686, pages 718-725, 2003.
[5] Z. Wang, H. Bao, L. Zhang, “PNN based motion detection with adaptive learning rate”, International Conference on Computational Intelligence and Security, CIS 2009, Beijing, December 2009.
[6] B. Do, S. Huang, “Dynamic Background Modeling based on Radial Basis Function Neural Networks for Moving Object Detection”, International Conference on Multimedia and Expo, ICME 2011, Barcelona, Spain, July 2011.
[7] G. Shobha, N. Satish Kumar, “Adaptive Background Modeling and Foreground Detection in Video Sequence Using Artificial Neural Network”, International Conference on Intelligent Computational Systems, ICICS 2012, Dubai, January 2012.
[8] J. Ramirez-Quintana, M. Chacon-Murguia, “Self-Organizing Retinotopic Maps Applied to Background Modeling for Dynamic Object Segmentation in Video Sequences”, International Joint Conference on Neural Networks, IJCNN 2013, August 2013.
[9] R. Athilingam, K. Kumar, G. Kavitha, “Neuronal mapped hybrid background segmentation for video object tracking”, International Conference on Computing, Electronics and Electrical Technologies, ICCEET 2012, pages 1061-1066, 2012.
[10] M. De Gregorio, M. Giordano, “Change Detection with Weightless Neural Networks", IEEE Change Detection Workshop, CDW 2014, June 2014.
[11] R. Guo, H. Qi, “Partially-Sparse Restricted Boltzmann Machine for Background Modeling and Subtraction”, International Conference on Machine Learning and Applications, ICMLA 2013, pages 209-214, December 2013.
[12] L. Xu, Y. Li, Y. Wang, E. Chen “Temporally Adaptive Restricted Boltzmann Machine for Background Modeling”, AAAI 2015, Austin, Texas USA, January 2015.
[13] C. Stauffer and E. Grimson, “Adaptive background mixture models for real-time tracking,” in IEEE Int. Conf. Comput. Vision and Pattern Recogn. (CVPR), vol. 2, (Fort Collins, Colorado, USA), pp. 246–252, June 1999.
[14] P.-L. St-Charles, G.-A. Bilodeau, and R. Bergevin, “A self-adjusting approach to change detection based on background word consensus,” in IEEE Winter Conf. Applicat. Comp. Vision (WACV), (Waikoloa Beach, Hawaii, USA), pp. 990–997, Jan. 2015.
[15] S. Bianco, G. Ciocca, and R. Schettini, “How far can you get by combining change detection algorithms?,” CoRR, vol. abs/1505.02921, 2015.
[16] A. Schick, M. Bauml, and R. Stiefelhagen, “Improving foreground segmentation with probabilistic superpixel Markov Random Fields,” in IEEE Int. Conf. Comput. Vision and Pattern Recog. Workshop (CVPRW), (Providence, Rhode Island, USA), pp. 27–31, June 2012.
[17] M. Sedky, M. Moniri, and C. Chibelushi, “Spectral 360: A physics-based technique for change detection,” in IEEE Int. Conf. Comput. Vision and Pattern Recog. Workshop (CVPRW), (Columbus, Ohio, USA), pp. 399–402, June 2014.
[18] P.-L. St-Charles, G.-A. Bilodeau, and R. Bergevin, “SuBSENSE: A universal change detection method with local adaptive sensitivity,” IEEE Trans. Image Process., vol. 24, pp. 359–373, Jan. 2015.
[19] A. Miron and A. Badii, “Change detection based on graph cuts,” in IEEE Int. Conf. Syst., Signals and Image Process. (IWSSIP), (London, United Kingdom), pp. 273–276, Sept. 2015.
[20] L. Maddalena and A. Petrosino, “The SOBS algorithm: what are the limits?,” in IEEE Int. Conf. Comput. Vision and Pattern Recog. Workshop (CVPRW), (Providence, Rhode Island, USA), pp. 21–26, June 2012.
[21] A. Elgammal, D. Harwood, and L. Davis, “Non-parametric model for background subtraction,” in European Conf. Comput. Vision (ECCV), vol. 1843 of Lecture Notes Comp. Sci., (Dublin, Ireland), pp. 751–767, Springer, June 2000.
[22] F. D. G. Allebosch, P. Veelart, and W. Philips, “EFIC: Edge based foreground background segmentation and interior classification for dynamic camera viewpoints,” in Advanced Concepts for Intelligent Vision Syst. (ACIVS), vol. 9386 of Lecture Notes Comp. Sci., pp. 130–141, Springer, Oct. 2015.