Markov Chain Models Based on the Genetic Algorithm for Texture Recognition

• Main Authors: You-Ping Huang, 黃有平
• Other Authors: none
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/66249085263091789558

碩士 === 立德管理學院 === 應用資訊研究所 === 93 === The Markov chain models (MCM) have been recently applied to many recognition applications. The well-known clustering algorithm, the k-means algorithm, is used to design the codebooks of the MCM, and then each codeword in the codebook is regarded as one state of the MCM. However, the users usually have no ideal to decide the number of states before the design of the MCM, and then the users doubt whether the MCM produced by the k-means algorithm is optimal or not. In this study, the new MCM based on the genetic algorithm is proposed for texture recognition. The genetic algorithm is the combination of the clustering algorithm and the design of the MCM, and it can automatically find the number of states in MCM. Thus, the users need not to define the size of codebook before the design of the MCM. Furthermore, the performance of the new MCM is emphasized to be as high as possible in the genetic algorithm. We have proposed the fuzzy Markov chain model and fuzzy genetic algorithm in addition to strengthen the of Markov chain model. Experimental results show that our proposed MCM achieves better performance than of the traditional MCM and other methods in texture recognition.