臺灣博碩士論文加值系統

在視訊監控中物件偵測區域的完整度是很重要的一件事(例如行人與車輛辨識)。然而一般的物件偵測技術中，因為目標物受容易受到顏色或亮度等變化所影響，使偵測的目標物區域不完整。為了克服這個問題，我們提出超像素迭代群組法精準的分割目標物的邊界並精準物件偵測中的目標物區域。使用超像素迭代群組法可以改善一般超像素分割不準確的問題，並保持目標物區域的完整性。我們認為每一塊超像素區域中偵測目標物像素越多，該超像素越有可能是目標物所在區域。因此，我們將偵測物件與超像素結合增加目標物區域的完整性。實驗結果顯示，使用超像素迭代群組精準物件可以比其它方法得到更完整的目標物區域。

The region completeness of object detection is very crucial to video surveillance, such as the pedestrian and vehicles identifications. However, many conventional object detecting techniques can’t guarantee the object region completeness because the object detections can be affected by lighting or color variations. In order to overcome this problem, we propose the iterative superpixels grouping (ISPG) method to extract the precise object boundary and generate the object region with high completeness after the object detections. By extending the superpixel segmentation method, the proposed ISPG method can improve the inaccurate segmentation problem and guarantee the region completeness on the object regions. The larger the number of object pixels in each superpixel, the more possibility this superpixel belongs to the object region. Hence, the region completeness can be improved by merging the superpixels that are classified into the object. Experimental results show that the proposed method outperforms the some conventional objection method in terms of object completeness evaluations.

目錄
摘要….....................................................................................................................................i
ABSTRACT….......................................................................................................................ii
目錄…...................................................................................................................................iv
圖目錄…................................................................................................................................v
表目錄…..............................................................................................................................vii
第一章 簡介..........................................................................................................................1
1.1 研究動機.........................................................................................................1
1.2 相關研究.........................................................................................................2
1.3 研究方法與貢獻.............................................................................................3
第二章 超像素迭代群組法..................................................................................................6
2.1影像的過分分割..............................................................................................7
2.2遞回標記法....................................................................................................13
2.3最佳配對法....................................................................................................15
2.4併吞合併法....................................................................................................17
第三章 ISPG之精準物件偵測技術...................................................................................20
第四章 實驗........................................................................................................................27
4.1測試影像與目標物之Ground-Truth建立....................................................27
4.2數據估算........................................................................................................29
4.3實驗數據與結果............................................................................................32
第五章 結論與未來發展....................................................................................................41
參考文獻..............................................................................................................................42



圖目錄
圖 1-1. 使用高斯模型建立背景模型出現物件破碎的情形。(a)原始圖(b) 用高斯模型所得到的物件偵測結果。…………………………………………………………………3
圖 1-2. Schick使用馬可夫隨機場建立機率超像素之物件偵測，因為目標物移動緩慢 出現物件破碎的情形。(a)SLIC超像素切割(b)PSP-MRF物件偵測結果。..................4
圖 1-3. 使用超像素迭代群組法之物件偵測系統流程圖。…………………………….5
圖 2-1. 超像素分割。…………………………………………………………………….6
圖 2-2. ISPG流程圖。…………………………………………………………………….7
圖 2-3. SLIC在不同尺寸的超像素影像分割。(a)將影像切成200個超像素。(b)將影像切成400個超像素。(c)將影像切成800個超像素。……………………………………8
圖 2-4. 減少超像素的搜索區域SLIC會制定出搜索區域，(a)是使用K-means演算法，群中心與像素的距離計算會涵蓋整張影像，(b)為SLIC方法，制定出搜索範圍，以減少群中心與像素的距離計算與迭代次數。………………………………………………11
圖 2-5. 影像執行過份分割法的結果。…………………..………………………………11
圖 2-6. SLIC系統流程圖。……………………………………………………………....12
圖 2-7. 使用遞迴標記法將特徵相似的超像素做結合的結果。……………………….13
圖 2-8. 遞迴標記法的系統流程圖。…………………………………….…………….…14
圖 2-9. 最佳配對法圖示。..................................................................................................16
圖 2-10. 使用最佳配對法合併的結果。…………………………………………………16
圖 2-11. 最佳配對法系統流程圖。………………………………………………………17
圖 2-12. 併吞合併法流程圖。……………………………………………………………18
圖 2-13. 併吞合併法執行結果。…………………………………………………………18
圖 2-14. 使用ISPG在自然影像中切割超像素的結果。………………………………19
圖 3-1. 使用GMM方法進行目標物偵測，(a)原始影像。(b)偵測影像。………………20
圖 3-2. 將單一像素的前景偵測結合ISPG方法之系統流程圖。……………………..21

圖 3-3. 使用過分分割法切割的超像素補償過度破碎的目標物。(a)原始影像分割後的超像素分布。(b)使用GMM物件偵測的結果。(c)分割後的超像素對應GMM物件偵測。(d)補償結果。………………………………………………………………………..22
圖 3-4. 使用ISPG方法將超像素合併，超像素內的區域完整度提高，利用區域內像素的關係性改善目標物破碎的問題。圖(a)分割後的超像素進行ISPG合併後的結果。圖(b)將第一階段中目標物補償的結果與ISPG的結果對應。圖(c)將目標物像素在超像素區域中的比例大於臨界值Tb ，該超像素區塊便被定義為超像素區塊。………….23
圖 3-5. 物件偵測與ISPG方法完成三階段結合的結果。(a)原圖。(b) 1010oversegmentation分割的結果。(c) 方法ISPG將超像素合併。(d) 使用GMM物件偵測的結果。(e) 使用圖b與物件偵測結合得到補償過後的物件。(f) 使用圖c與物件偵測結合得到補償過後的物件。(g)為圖e加圖f，灰色地帶為補償的地區。(h) 物件偵測與ISPG方法完成三階段結合。………………………………………………….24
圖 3-6. 目標物大幅度破碎在不同臨界值下的物件補償結果(a)原始影像。(b)GMM物件偵測。(c)~(j)臨界值大小為0.1~0.9之物件補償的結果。臨界值越小，目標物破碎的部分容易越容易補償大幅度破碎的部分。……………………………………………25
圖 3-7. 影像中背景出現雜點在不同臨界值下的物件補償結果(a)原始影像。(b) GMM物件偵測。(c)~(j)臨界值大小為0.1~0.9之物件補償的結果。臨界值越小，目標物容易受到背景雜點所影響，將背景誤判成目標物。………………………………………26
圖 4-1. CDnet提供的六類不同環境條件測試影像。(a)基線。 (b)動態背景。 (c)相機抖動。 (d)陰影。 (e) 間歇性的目標物移動。 (f)熱感應。………….……………………28
圖 4-2. Ground-Truth中，五種不同的灰階代表的區域。..............................................29
圖 4-3. Baseline類別中pedestrians影像的執行結果，(a)~(f)原始影像。(g)~(l) GMMkaew物件偵測的結果。(m)~(r) GMMkaew+ISPG執行物件偵測結果。………………………34
圖 4-4. cameraJitter類別中boulevard影像的執行結果，(a)~(f)原始影像。(g)~(l) GMMkaew物件偵測的結果。(m)~(r) GMMkaew+ISPG執行物件偵測結果。比較(l)與(r)利用ISPG去掉背景中細碎的雜點。…………………………………………………….36
圖 4-5. dynamicBackground類別中overpass的執行結果，(a)~(f)原始影像。(g)~(l) GMMkaew物件偵測的結果。(m)~(r) GMMkaew+ISPG執行物件偵測結果。……………37
圖 4-6. shadow類別中peopleInShade的執行結果，(a)~(f)原始影像。(g)~(l) GMMkaew物件偵測的結果。(m)~(r) GMMkaew+ISPG執行物件偵測結果。……………………….38
圖 4-7. thermal類別中library的執行結果，(a)~(f)原始影像。(g)~(l) GMMkaew物件偵測的結果。(m)~(r) GMMkaew+ISPG執行物件偵測結果。……………………………….40




表目錄
Algorithm 1. SLIC超像素分割。…………………...………….…………………………11
Algorithm 2. 遞回標記法。...............................................................................................14
Algorithm 3: 最佳匹配法。………………………………………………………………16
Algorithm 4: 併吞合併法。……………………………………………………………...19
Table 1.物件影像與Ground-Truth比對分析表。………………………………………...30
Table 2. 使用Change Detection方法估算GMMkaew、GMMkaew+PSP-MRF、GMMkaew
+ISPG之數據結果。………………………………………..………………………...31
Table 3. 使用Change Detection方法估算SOBS、SOBS+PSP-MRF、SOBS+ISPG之數據結果。………………………………………..………………………..…………….…...32

參考文獻
[1] C. C. Lien, J. C. Wang and Y. M. Jiang, “Multi-mode Target Tracking on a Crowd Scene,” Proceedings of the Third International Conference on International Information Hiding and Multimedia Signal Processing, vol. 02, pp. 427-430, 2007.
[2] A. Elgammal, D. Harwood and L. Davis, “Non-parametric model for background subtraction,” in proceedings of the 6th European Conference on Computer Vision, pp. 751-767,2000.
[3] C.Stauffer, and W. E. L. Grimson, “Adaptive background mixture models for real-time tracking,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition, vol. 02, 1999.
[4] Zivkovic, Z. and F. van der Heijden, “Recursive unsupervised learning of finite mixture models,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 26, pp. 651-656, 2004.
[5] H. H. Lin, J. H. Chuang, T. L. Liu “Recursive unsupervised learning of finite mixture models” IEEE Transactions on Image Processing, vol. 20, pp. 822-836, 2011.
[6] C. C. Lien, K. L. Yu. “Blur Image Segmentation using Iterative Super Pixels Grouping Method”, 2012.
[7] P. KaewTraKulPong, and R. Bowden, “An improved adaptive background mixture model for real-time tracking with shadow detection,” Computer Vision and Distributed Processing, pp. 135-144, 2002.
[8] X. Dai and S. Khorram. “Performance of Optical Flow Techniques, ”IEEE Computer Society Conference on Computer Vision and Pattern Recognitions, pp. 236-242, 1992.
[9] S. Wang, X. Wang, and H. Chen, “A stereo video segmentation algorithm combining disparity map and frame difference,” International Conference on Intelligent System and Knowledge Engineering, vol.1, pp.1121-1124, 2008.
[10] K. Kim, T. H. Chalidabhongse, D. Harwood, and L. S. Davis, “Real-Time Foreground
-Background Segmentation using Codebook Model”, Real-Time Imaging, pp.172-185, 2005.
[11] K.S. Devi, N.Malmurugan, S.Poornima, “Improving the Efficiency of Background Subtraction using Super pixel Extraction and Midpoint for Centroid” International Journal of Computer Applications, vol. 43, pp. 0975-8887, 2012.
[12] S. Wang, H. Lu, F.Yang and M. H. Yang, “Superpixel tracking,” IEEE International Conference on Computer Vision (ICCV), pp. 1323-1330, 2011.
[13] R. Achanta, A. Shaji, K. Smith, A. Lucchi, P. Fua and S. Suutrunk, "SLIC Superpixels Compared to State-of-the-art Superpixel Methods," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 34 pp.2274-2282, 2011.
[14] A. Schick, M. Baumi, R. Stiefelhagen, “Improving foreground segmentations with probabilistic superpixel Markov random fields,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pp. 27-31, 2012.
[15] http://en.wikipedia.org/wiki/Markov_random_field
[16] Xi Li and H. Sahbi, “Superpixel-based object class segmentation using conditional random fields,” IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 1101-1104, 2011.
[17] http://en.wikipedia.org/wiki/Conditional_random_field
[18] C. Liu and Z. Zhao, “Person Re-identification by Local Feature Based on Super Pixel,” Advances in Multimedia Modeling Lecture Notes in Computer Science, vol. 7732, pp 196-205, 2013.
[19] http://en.wikipedia.org/wiki/Lab_color_space
[20] N. Goyrtte, P. Jodoin, F. Porikli and J. Konrad, “Changedetection.net: A new change detection benchmark dataset,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pp. 1-8, 2012.
[21] P. KaewTraKulPong and R. Bowden, “An improved adaptive background mixture model for realtime tracking with shadow detection,” EuropeanWorkshop on Advanced Video Based Surveillance Systems, pp. 135-144, 2001.
[22] L. Maddalena and A. Petrosino, “The SOBS algorithm: what are the limits?,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) ,pp. 21-26, 2012.