臺灣博碩士論文加值系統

在以視覺為基礎的人物定位研究中，人物定位是一個重要且具挑戰性的研究課題。基於機率占有率圖(probabilistic occupancy map，POM)人物定位方法因爲能在嚴重的遮蔽情況下得到精準的人物定位結果，所以成爲人物定位方法的主流之一。而POM在人物定位方法的計算複雜度非常高，過程會耗費許多時間。本論文在改良過後的POM 方法上，再提出了三種不同的人物預測位置收斂方法，希望在不降低定位準確度的情況之下，提升POM的計算效率。本論文提出三個預測收斂方法進行改進：(i)到N收斂回合就將剩餘未判定位置收斂；(ii)剩餘N_1個位置須判定則採用新的佔有率機率停止條件；(iii)參考四周位置是否有人物存在，若是已有人物存在則目前位置也會判定為有人存在。本論文測試不同機率收斂方法於原圖影像與基於線段取樣之壓縮影像的定位結果，實驗結果顯示，提出的方法能提升基於機率占有率圖人物定位方法的速度，並且人物定位的精準度也能維持在一定水準。

People localization is an important and challenging research topic with the vision-based security surveillance systems, the improvement in accuracy and efficiency of people localization has attracted remarkable research efforts. Recently, employing probabilistic occupancy map (POM) becomes one of the main research trends in people localization due to its great localization accuracy under severe occlusions and adverse lighting conditions. However, the computation complexity of the POM-based approach is high. We expect to improve the computational efficiency of the approach without reducing much localization accuracy. Based on the improved POM method, three enhanced convergence methods are proposed in this thesis: (i) Converge the remaining undecided potential people locations immediately after the N iterations, and (ii) Converge the rest N1 potential people locations by adopting new termination condition for occupancy probability estimation, and (iii) The location is regarded as occupied if it is surrounded by people. In this paper, we compare the people detection results of the three enhanced convergence methods combined with the original images and the condensed images obtained by line sampling. Experimental results show that the proposed method can improve the efficiency of the POM-based approach while still maintaining localization accuracy.

摘 要 I
Abstract II
誌 謝 III
目 錄 IV
圖目錄 VI
表目錄 X
第一章 簡介 1
1.1 研究動機與背景 1
1.2 相關研究與文獻縱覽 2
1.2.1 單攝影機偵測人物方法 2
1.2.2 多攝影機偵測人物方法 4
第二章 基於機率佔有率人物定位方法- POM (probabilistic occupancy map) 7
2.1 機率佔有率人物定位簡介 7
2.2 基於機率佔有率人物定位方法的改良 9
2.3 機率佔有率人物定位收斂方法改進 12
2.3.1 候選網格位置(candidate grid locations，CGLs) 12
2.3.2 候選網格位置上虛擬人型立板的機率收斂計算 13
第三章 加速機率佔有率人物定位之機率收斂方法 15
3.1 原始影像機率資料 15
3.1.1 虛擬人型立板機率收斂 15
3.1.2 每回合機率收斂後人形立板下降數目 24
3.1.3 人形立板收斂情形 28
3.2 機率收斂方法 30
3.2.1 方法I — 迴圈數超過定值後的收斂方法 30
3.2.2 方法II — 剩餘N1個虛擬人形立板以不同方法收斂 31
3.2.3 方法III — 參考四周候選網格收斂方法 31
第四章 實驗結果 33
4.1 實驗環境 33
4.1.1 實驗所用設備與測試資料介紹 33
4.1.2 實驗所用方法與介紹 33
4.2 GVP轉換後的影像與收斂方法合併之結果 34
4.3 合併壓縮影像與收斂方法合併結果 35
4.3.1 影像解析度為120×240之結果 35
4.3.2 影像解析度為360×48之結果 37
4.3.3 影像解析度為180×48之結果 39
4.3.4 影像解析度為120×48之結果 41
4.3.5 壓縮21倍、14倍、9倍、6倍、4倍與原比例之實驗結果 43
4.4 使用GVP影像測試不同閥值之實驗結果 46
4.4.1 方法I使用不同閥值之實驗結果 46
4.4.2 方法II使用不同閥值之實驗結果 47
第五章 結論 49
參考文獻 50
附錄-壓縮影像相關資料分析 52
A.1 壓縮影像資料 52
A.2 壓縮影像每回合機率收斂後人形立板下降數目 53
A.3 壓縮影像人形立板收斂情形 57

[1] N. Dalal and B. Triggs, “Histograms of oriented gradients for human detection,” IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2005.
[2] Y. Jiang, J. Ma, "Combination features and models for human detection", IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2015.
[3] J. Fan, W. Xu, Y. Wu, and Y. Gong, “Human tracking using convolutional neural networks,” IEEE Trans., Neural Netw., vol. 21, no. 10, pp. 1610–1623, Oct. 2010.
[4] M. Everingham, L. Van Gool, C. K. I. Williams, J. Winn, and A. Zisserman. The PASCAL Visual Object Classes Challenge 2010 (VOC2010) Results. http://www.pascalnetwork.org/challenges/VOC/voc2010/workshop/index.html.
[5] P. Felzenszwalb, R. Girshick, D. McAllester, and D. Ramanan, “Object detection with discriminatively trained part based models,” IEEE Trans., Pattern Anal. Machine Intell., vol.32, no. 9, pp. 1627–1645, Sept. 2010.
[6] P. Felzenszwalb and D. McAllester, “Object detection grammars,” IEEE Int. Conf. Comput. Vis. Workshops, 2011.
[7] R. Girshick, P. Felzenszwalb, and D. McAllester, “Object detection with grammar models,” Neural Inf. Proces. Syst. (NIPS), 2011.
[8] G. Shu, A. Dehghan, O. Oreifej, E. Hand, and M. Shah, “Part-based multiple-person tracking with partial occlusion handling,” IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2012
[9] T.-H. Chang and S. Gong, “Tracking multiple people with a multicamera system,” IEEE Workshop Multi-Object Tracking, 2001.
[10] A. Mittal and L. Davis, “M2 Tracker: a multi-view approach to segmenting and tracking people in a cluttered scene,” Int. J. Comput. Vis., vol. 51, no. 3, pp. 189–203, Feb. 2003.
[11] U. Akos and B. Csaba. “A bayesian approach on people localization in multicamera systems,” IEEE Trans., Circuits Syst. Video Technol., vol.23, no.1 , Jan. 2013.
[12] Q. Zhang and K. N. Ngan, “Segmentation and tracking multiple objects under occlusion from multiview video,” IEEE Trans., Image Process., vol. 20, no. 11, pp. 3308–3313, Nov. 2011.
[13] T.-E. Tseng, A.-S. Liu, P.-H. Hsiao, C.-M. Huang and L.-C. Fu. “Real-time people detection and tracking for indoor surveillance using multiple top-view depth cameras,” IEEE/RSJ Int. Conf. Intell. Robots Syst.(IROS), 2014.
[14] A. Elfes, “Using occupancy grids for mobile robot perception and navigation,” IEEE Comput., vol. 22, no. 6, pp. 46–57, Jun. 1989.
[15] F. Fleuret, J. Berclaz, R. Lengagne, and P. Fua, “Multicamera people tracking with a probabilistic occupancy map,” IEEE Trans., Pattern Anal. Machine Intell., vol. 30, no. 2, pp. 267–282, Feb. 2008.
[16] Y.-S. Lin, K.-H. Lo, H.-T. Chen, and J.-H. Chuang, “Vanishing Point-Based image transforms for enhancement of probabilistic occupancy map-based people localization,” IEEE Trans., Image Process., vol. 23, no. 12, pp. 5586-5598, 2014.
[17] Y.-S. Lin, K.-H. Lo, H.-T. Chen, and J.-H. Chuang, “Enhancement of probabilistic occupancy map-based people localization,” Doctoral dissertation, National Chiao Tung University, 2016.
[18] Y.-S. Lin, H.-T. Chen, and J.-H. Chuang, “An efficient probabilistic occupancy map-based people localization approach,” IEEE Int. Conf. Visual Comm. Image Process., 2015.
[19] K.-H. Lo and J.-H. Chuang, “Vanishing point-based line sampling for real-time people localization,” IEEE Trans., Circuits Syst. Video Technol., vol. 23, no. 7, pp. 1209–1223, Jul., 2013.