隨著攝影機普及以及網路頻寬的不斷升級，監控系統(video surveillance system)愈來愈被人們所重視。行人辨識(person re-identification)是監控領域中最為核心的部分，辨識出在不同攝影機視角、光照變化及人身少部分遮蔽物情況下的目標人物。目前有許多方法都有不錯的正確率，但這些方法所使用的特徵擷取/特徵比對方法過於耗時，導致整個系統難以有效實際應用。
本論文提出兩套以顏色特徵為基礎並有別於現有方法的高效能單張影像行人辨識系統，分別使用高斯混合模型(Gaussian Mixture Model)結合低層級(low-level)與中層級(mid-level)的顏色特徵來提高系統辨識率及運算效能。兩種系統架構皆分為3個主層級：第一層級為光源正規化(illumination normalization)，用來移除環境光源分佈不均的影響；第二層級包含背景/前景分割以及人體部位切割：因為所處的環境太複雜會影響特徵點截取及比對，透過此步驟事先過濾冗餘資訊並將人身的各部位切割出來，能使後續取特徵點權重分佈之步驟更為精準，有效提高系統的正確率；第三層級為特徵點的截取及比對。在第一個系統GMMWCH中所使用的低層級特徵點為顏色直方圖(color histogram)，能在保持高辨識率情況下同時擁有高速的運算效能；而第二個系統SaliGMMWCH使用中層級的斑紋特徵點(patch features)，因為此特徵之計算複雜度較顏色直方圖高，故本系統的運算效能會比GMMWCH差，取而代之的是擁有更為準確的辨識率，在此也利用自定義的權重篩選來合併不同特徵方法以進一步提高辨識率。
本論文所提出的兩種系統皆在僅使用單張影像情況下就能辨識出在不同攝影機視角、受環境光源變化及少部分遮蔽物人身的目標人物，透過實驗結果也顯示出本論文所提理論的準確率及效能皆比過去代表性方法有更卓越的進步。

With the development of electronics in recent years, more and more people started to concern with the issue of video surveillance and it has been widely used for various purposes, such as public safety, privacy protection, facilities surveillance and traffic monitoring. One of these issues is that matching pedestrian over different non-overlapping camera views, known as the person re-identification problem. It is a novel and challenging research topic in computer vision due to the large illumination variations, visual appearance changes caused by different viewpoints and partial occlusions. Although there are a lot of existing methods have good recognition rate, they take too much time for feature extracting and matching so that the system cannot be used in the real application.
In order to deal with these challenges, we propose two different and efficient color-based-method for the single-shot person re-identification, which uses the Gaussian mixture model to combine with color histograms (low-level feature) and the dense salient patches (mid-level feature) as the color features. Both the two proposed systems are three-stage processes. The first stage is the image enhancement by illumination normalization, because the proposed frameworks are based on the color histograms combined with spatial information of body segments, the intensity variations plays a key role in the matching rate. The second stage includes pedestrian segmentation and human region partition. That separates the background (BG) and foreground (FG) and locate the body segments can improve the accuracy of feature extracting and matching. The third stage is used to perform feature extracting and matching. For the first system, GMMWCH, we employ the Gaussian mixture model to generate the weighted color histogram as the color features. It can provide low computation and good recognition rate. For the second system, SaliGMMWCH, we employ the dense correspondence to link the color histogram weighted by the Gaussian mixture model to find salient regions. Even though that takes more time for computation, the SaliGMMWCH retains a better recognition rate than GMMWCH. In addition, we choose the correct match by matching the similarity scores of different feature with appropriate weight selection.
Both our proposed approaches for the single-shot person re-identification have been tested on the public benchmark, VIPeR, for evaluation. The experimental results demonstrate superior recognition rate and execution performance by using our proposed methods compared to the other representative methods.

[1] N. Jojic, A. Perina, M. Cristani, V. Murino, and B. Frey. “Stel component analysis: Modeling spatial correlations in image class structure,” in Computer Vision and Pattern Recognition, 2009, pp. 2044-2051.
[2] M. Farenzena, L. Bazzani, A. Perina, V. Murino, and M. Cristani, “Person re-identification by symmetry-driven accumulation of local features,” in Computer Vision and Pattern Recognition, 2010, pp. 2360-2367.
[3] S. Byers and A. Raftery, “Nearest-neighbor clutter removal for estimating features in spatial point processes,” in Journal of the American Statistical Association, pp. 577-584, Jun 1998.
[4] Y. Du, H. Ai, S. Lao. “Evaluation of Color Spaces for Person Re-identification,” in International Conference on Pattern Recognition, 2012, pp. 1371-1374.
[5] D. Gray and H. Tao, “Viewpoint invariant pedestrian recognition with an ensamble of localized features,” in European Conference on Computer Vision, 2008, pp. 262-275.
[6] S. Bak, E. Corvee, F. Br’emond, M. Thonnat, “Person re- identification using Haar-based and DCD-based signature,” in Workshop on Activity Monitoring by Multi-Camera Surveillance Systems , 2010.
[7] B. Prosser, W.-S. Zheng, S. Gong, and T. Xiang. “Person re-identification by support vector ranking,” in British Machine Vision Conference, 2010.
[8] M. Hirzer, C. Beleznai, P. Roth, and H. Bischof. “Person re-identification by descriptive and discriminative classification,” Image Analysis, 2011.
[9] W. Schwartz and L. Davis, “Learning discriminative appearance-based models using partial least squares,” in XXII Brazilian Symposium on Computer Graphics and Image Processing (SIBGRAPI), 2009.
[10] Y.-C. Chang, C.-K. Chiang, S.-H. Lai, “Single-Shot Person Re-identification Based on Improved Random-Walk Pedestrian Segmentation,” in Intelligent Signal Processing and Communication Systems, 2012.
[11] R. Zhao, W. Ouyang, and X. Wang, “Unsupervised salience learning for person re-identification,” in Computer Vision and Pattern Recognition, 2013.
[12] Lowe D., “Distinctive Image Features from Scale-Invariant Keypoints,” International Journal of Computer Vision, Vol. 60, pp. 91-110, Springer, 2004.
[13] B. Ma, Y. Su, and F. Jurie, “ Local descriptors encoded by fisher vectors for person re-identification,” in European Conference on Computer Vision, 2012.
[14] B. Ma, Y. Su, and F. Jurie, “Bicov: a novel image representation for person re-identification and face verification,” 2012.
[15] X. Wang, G. Doretto, T. Sebastian, J. Rittscher, and P. Tu, “Shape and appearance context modeling,” in International Conference on Computer Vision, 2007.
[16] N. Gheissari, T. Sebastian, and R. Hartley, “Person re-identification using spatiotemporal appearance,” in Computer Vision and Pattern Recognition, 2006.
[17] S. Bak, E. Corvee, F. Br’emond, M. Thonnat, “Person re-identification using spatial covariance regions of human body parts,” in Advanced Video and Signal-Based Surveillance, 2010.
[18] D. Cheng, M. Cristani, M. Stoppa, L. Bazzani, and V. Murino, “Custom pictorial structures for re-identification,” in British Machine Vision Conference, 2011.
[19] W-S. Zheng, S. Gong, and T. Xiang, “Person re-identification by probabilistic relative distance comparison,” in Computer Vision and Pattern Recognition, 2011.
[20] D. Gray, S. Brennan, and H. Tao, “Evaluating appearance models for recognition, reacquisition, and tracking,” in 10th IEEE International Workshop on Performance Evaluation of Tracking and Surveillance (PETS), 09/2007 2007.
[21] C. Liu, S. Gong, C. Loy, and X. Lin, “Person re-identification: What features are important? ,” in European Conference on Computer Vision, 2012.
[22] K. L. M. Cho, “Bilateral symmetry detection and segmentation via symmetry-growing,” in British Machine Vision Conference, 2009.
[23] Bhattacharyya, A. “On a measure of divergence between two statistical populations defined by their probability distributions,”
Bulletin of the Calcutta Mathematical Society 35: 99-109. MR 0010358, 1943.
[24] K. Ma and J. Ben-Arie, “Vector array based multi-view face detection with compound exemplars,” in Computer Vision and Pattern Recognition, 2012.
[25] C. Liu, J. Yuen, and A. Torralba, “Sift flow: Dense correspondence across scenes and its applications,” in Transactions on Pattern Analysis and Machine Intelligence, 2011.