Application of Case-Based Reasoning and Neural Network on the Image Classification

• Main Authors: Hung-Hsu Chang, 張宏旭
• Other Authors: Alan Liu
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/11964610762555525353

碩士 === 國立中正大學 === 電機工程研究所 === 93 === The research of Image Classification has been applied to many domains, like medical science, security, business etc., and become a powerful assistance. And, because of the non-linear, high-complex, and distributed properties, Neural Network is a popular application in Image Classification. However, it would take time and effort on the design of Neural Network architectures to achieve a good ability of generalization. A cased-based reasoning (CBR) system makes use of past experience to solve the new problems and is capable of the incremental learning. If we can take the advantage of the CBR property of incremental learning to enhance the generalization of Neural Network, then we can avoid time-consuming of trial and error. The approach proposed in this thesis exploits a trained Neural Network in some environmental conditions and stores its good result in the case base for reusing. Then, by the process of CBR, classification is tested generally to know whether it works or not. Meanwhile, by applying the trained Neural Network to the procedure of CBR process, we can determine whether the retrieved case should be retained into the case base or not, and make it to be an automatic system.