The Hierarchical Clustering Analysis of Hyperspectral Image Based on Probabilistic Latent Semantic Analysis
YI Wen-bin1,2, SHEN Li1,2, QI Yin-feng1,3, TANG Hong1,4,5*
1. State Key Laboratory of Earth Surface Processes and Resource Ecology, Beijing Normal University, Beijing 100875, China 2. College Institute of Resources Science & Technology, Beijing Normal University, Beijing 100875, China 3. China University of Mining and Technology, Key Laboratory for Land Environment and Disaster Monitoring of SBSM, Xuzhou 221116, China 4. Key Laboratory of Mine Spatial Information Technologies, State Bureau of Surveying and Mapping, Jiaozuo 454010, China 5. Key Laboratory of Environmental Change and Natural Disaster, Ministry of Education, Beijing Normal University, Beijing 100875, China
Abstract：The paper introduces the Probabilistic Latent Semantic Analysis (PLSA) to the image clustering and an effective image clustering algorithm using the semantic information from PLSA is proposed which is used for hyperspectral images. Firstly, the ISODATA algorithm is used to obtain the initial clustering result of hyperspectral image and the clusters of the initial clustering result are considered as the visual words of the PLSA. Secondly, the object-oriented image segmentation algorithm is used to partition the hyperspectral image and segments with relatively pure pixels are regarded as documents in PLSA. Thirdly, a variety of identification methods which can estimate the best number of cluster centers is combined to get the number of latent semantic topics. Then the conditional distributions of visual words in topics and the mixtures of topics in different documents are estimated by using PLSA. Finally, the conditional probabilistic of latent semantic topics are distinguished using statistical pattern recognition method, the topic type for each visual in each document will be given and the clustering result of hyperspectral image are then achieved. Experimental results show the clusters of the proposed algorithm are better than K-MEANS and ISODATA in terms of object-oriented property and the clustering result is closer to the distribution of real spatial distribution of surface.
Key words：Probabilistic latent semantic analysis;Hyperspectral image;Hierarchical clustering;Semantic information
[1] LI Fei, ZHOU Chen-hu, CHEN Rong-guo(李 飞, 周成虎, 陈荣国). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2008, 28(11): 2482. [2] HU Xin-tang, ZHANG Bin, LI Jun-sheng, et al(胡兴堂, 张 兵, 李俊生, 等). Journal of Remote Sensing(遥感学报), 2005, 9(5): 604. [3] Xia Gui-song, He Chu, Sun Hong. IEEE Geoscience and Remote Sensing Letters, 2007, 4(4): 596. [4] Aguera F, Aguilar F J, Aguilar M A. ISPRS Journal of Photogrammetry & Remote Sensing, 2008, 63(6): 635. [5] Benz U C, Hofmann P, Willhauck G, et al. Photogrammetry & Remote Sensing, 2004, 58(3-4): 239. [6] CHEN Yun-hao, FENG Tong, SHI Pei-jun, et al(陈云浩, 冯 通, 史培军, 等). Wuhan University Journal·Natural Science Edition(武汉大学学报·信息科学版), 2006, 31(4): 316. [7] Poggi G, Scarpa G, Zerubia J B. IEEE Transactions On Geoscience And Remote Sensing, 2005, 43(8): 1901. [8] Laliberte A S, Rango A, Havstad K M, et al. Remote Sensing of Environment, 2004, 93: 198. [9] Akcay H G, Aksoy S. IEEE Transactions on Geoscience and Remote Sensing, 2008, 46(7): 2097. [10] Landauer T, Mcnamara D S, Dennis S, et al. Handbook of Latent Semantic Analysis. Hillsdale: Lawrence Erlbaum Associates, 2007. 427. [11] Quelhas P, Monay F, Odobez J M, et al. Proceedings of the 10th International Conference on Computer Vision，2005. [12] Li Fei-fei, Perona P. Proceeding of IEEE Conference on Computer Vision and Pattern Recognition，2005. [13] Barnard K, Duygulu P, Forsyth D, et al. Journal of Machine Learning Research. 2003, 3: 1107. [14] ZHOU Hui, GUO Jun, ZHU Chang-ren, et al(周 晖，郭 军，朱长仁，等). Journal of Remote Sensing (遥感学报), 2010, 14(4): 672. [15] SHI Jing, DAI Guo-zhong(石 晶，戴国忠). Journal of Computer Research and Development(计算机研究与发展), 2007, 44(2): 242. [16] Hofmann T. Machine Learning，2001, 42(1): 177. [17] Lienou M, Maitre H, Datcu M. IEEE Geoscience and Remote Sensing Letters, 2010, 7(1): 28. [18] Kyrgyzov I O, Kyrgyzov O O, Maitre H, et al. Proceedings of the 5th International Conference on Machine Learning and Data Mining in Pattern Recognition, 2007. [19] Frohn R C, Hao Y. Remote Sensing of Environment, 2006, 100(2): 237.
