智能化异源图像自适应融合体系的研究

项目名称： 智能化异源图像自适应融合体系的研究

项目编号： No.60874106

项目类型： 面上项目

立项/批准年度： 2009

项目学科： 金属学与金属工艺

项目作者： 赵美蓉

作者单位： 天津大学

项目金额： 30万元

中文摘要： 本项目密切跟踪图像融合的最新算法和规则，建立了完整的红外与可见光图像自适应融合理论框架；搭建了基于PC机和基于FPGA的图像融合硬件系统；提出了基于K-L变换的自动加权融合算法，该方法既保持了现有的加权平均融合算法的快速实用性，又将源图像中的有用信息集中到相互独立的新主成分矩阵中，并由此确定加权平均融合方法中源图像的加权系数；深入研究了基于变换域图像融合算法，提出了基于NSCT和PCNN的异源图像融合方法以及基于提升方向波的异源图像融合算法；在PC机或FPGA中实现了上述图像融合算法，获得了很好的融合效果；建立了基于模糊神经网络的融合效果综合评价体系，以便于对融合图像进行效果评价；最后将融合理论应用在一些实际的工程课题，如工件检测，复杂环境中尺寸测量等，获得了良好的效果。

中文关键词： 异源；图像融合；自适应；FPGA；模糊神经网络

英文摘要： The latest image fusion algorithms and rules are closely tracked in this project. A complete adaptive theoretical fusion framework of infrared and visible images is established; PC-based and FPGA-based hardware image fusion systems were built; K-L transform based automatic weight fusion algorithm is proposed, which not only maintains existing weight fusion algorithms' rapidity and practicality, but also concentrates useful information of source images into new mutually independent main Component matrix and determine weight coefficient of source images; PCNN-based and NSCT-based heterogeneous image fusion, and enhanced direction wave based image fusion algorithm are proposed by in-depth study of the image fusion algorithm based on transform domain；The above-mentioned image fusion algorithms are realized on PC and FPGA with good fusion effect. A comprehensive fusion evaluation system is established based on fuzzy neural network in order to evaluate the effect of the fusion image. Finally, this fusion theory is applied quite well in a number of practical engineering issues, such as parts inspection, dimensional measurement in complex environments and so on.

英文关键词： heterology; image fusion; self-adaption; FPGA; FNN(fuzzy neural network)