【基於图像信息的模式识别方法与应用研究】李炜.pdf

ADissertationSubmittedinPartial FulfillmentoftheRequirements fortheDegreeofDoctorofPhilosophyinEngineering Research inMethod and Application of Pattern Recognition BasedonImageInformation Ph.D.Candidate:LiWei Major:ControlTheoryandControlEngineering Supervisor:Prof.

华中科技大学博士学位论文分类器,将多特征的模式识别问题转化为不同低维特征空间的划分与融合问题,并具备了较强的学习能力和适应性。该分类器在保证识别速度的同时,可有效提高识别率,在学习能力、适应性以及可移植性方面亦表现较好.详细介绍了基于模糊神经网络的子集成分类器的原理、结构和设计方法,分类器集成策略采用基于监督学习的神经网络,以人机结合的方式,把三个模糊神经网络模型得到的识别结果加以集成,充分发挥各自的优势,进而获得比单个分类器更好的识别效果。

华中科技大学博士学位论文 sub-classifiers,including1level classifier based on expert system with a full use of expert experiences,2nd level classifier based on decision tree which turns a complex multi-class patternrecognition probleminto several simple ones,and3dlevel classifier basedonintegratedfuzzyneuralnetw