BP神经网络在多源图像融合中的应用研究  

Application Research of BP Neural Network in the Multi-source Image Fusion

作　　者：徐金龙[1] 庞尚珍[1] 

机构地区：[1]四川理工学院自动化与电子信息学院,四川自贡643000

出　　处：《四川理工学院学报（自然科学版）》2010年第6期707-711,共5页Journal of Sichuan University of Science & Engineering(Natural Science Edition)

基　　金：国家863高技术研究发展计划项目(2008AA11A134);四川理工学院人才引进科研启动项目(07ZR25)

摘　　要：采用BP神经网络与SVR融合法相结合的方法来实现像素级多源图像融合。改进型的SVR方法,用来进行全色波段和多光谱波段的融合。首先将各多光谱波段与全色波段求算回归系数,然后利用回归系数与多光谱波段合成模拟高空间分辨率全色波段,最后利用比值变换完成各波段的融合。通过仿真证明,BPNN-SVR融合法优于SVR融合法。The method which unified the BP neural network and the SVR fusion law to realize the pixel-leve 1 muhiple source image fusion is used in this paper. The advanced version SVR method was used to carry on the panchromatic wave band and the multi-spectrum wave hand fusion. First asks each multi-spectrum wave band and the panchromatic wave band to calculate the regression coefficient, then simulates the high spatial resolution panchromatic wave band using the regression coefficient with the multi-spectrum wave band synthesis, finally completes various wave bands using the ratio transformation the fusion. Proved through the simulation that, .the BPNN-SVR fusion method surpasses the SVR fusion method.

关 键 词：像素级 图像融合 BP神经网络 SVR方法 

分 类 号：TN751[电子电信—电路与系统]