Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/7294
標題: 群集技術在影像分割及資料分析上的應用
Image segmentation and data analysis using clustering techniques
作者: 葉佳威
yeh, gai-wei
關鍵字: clustering analysis;群集分析;microarray;Image segmentation;微陣列;影像分割
出版社: 電機工程學系
摘要: 
本文提出了不同的群集分析方法的應用和不同領域中進行群集分析的流程。本文一開始回顧了一些較常用的群集演算法,包括了階層式群集演算法和分割式群集演算法,其中分割式群集演算法包括有模糊均值演算法,可能性均值演算法。接下來,我們介紹一個最近提出的強健性演算法叫相似度演算法,這個演算法將會表現出強健性的特性。
群集演算法的應用主要包括兩個部分,微陣列的分析和影像分割。微陣列分析包括如何對資料進行預處理,如何使用四種不同的群集演算法進行群集分析包括階層式演算法,模糊均值演算法,可能性均值演算法,相似度演算法等。並且比較不同的預處理方法和不同的群集分析的方法。影像分割包括不同特徵向量的使用,群集分析,二值化,後處理和偵測出影像邊界。
本文最後我們提出群集分析未來的發展方向,並對不同領域的群集分析提出不同的建議,以期能夠增進群集分析的效能和優點。

This research presents the application of different clustering methods and the procedure of each application in clustering analysis. The research first previews some common clustering algorithm such as hierarchical clustering methods and partitional clustering methods which include FCM and PCM. Also, we have introduced a recently presented robust clustering method, named similarity-based clustering method. This clustering method will show majority of robustness.
The application in clustering methods includes two aspects: microarray analysis and image segmentation. The microarray analysis shows how to preprocess the data, how to use clustering analysis as following four clustering methods: FCM, PCM, and SCM. Furthermore, it compares different preprocessing methods and different clustering methods. The image segmentation includes usage of different feature vectors, clustering analysis, binary, post processing, and lastly the edge detection of the binary image.
Finally, we will make recommendations on the future of the clustering techniques, and provide different suggestions on clustering analysis in different applications, wishing to improve the efficiency and increase advantages.
URI: http://hdl.handle.net/11455/7294
Appears in Collections:電機工程學系所

Show full item record
 
TAIR Related Article

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.