利用改进T-S模糊神经网络恢复MMW图像  

MMW image restoration using modified T-S fuzzy neural network

作　　者：尚丽[1] 周燕[1] SHANG Li;ZHOU Yan(School of Electronic Information Engineering, Suzhou Vocational University, Suzhou 215104, China)

机构地区：[1]苏州市职业大学电子信息工程学院,江苏苏州215104

出　　处：《计算机工程与设计》2018年第5期1463-1466,1489,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61373098);江苏省自然科学基金青年基金项目(BK20160361)

摘　　要：为有效消除毫米波(MMW)图像中的非线性噪声,利用T-S模糊神经网络(T-S-FNN)对不确定信息进行有效区分的特性,实现MMW图像中非线性信息噪声的逼近,达到消噪的目的。为克服T-S-FNN规则冗余的缺点,考虑前件网络基于自适应模糊聚类的隶属度函数约束及后件网络的权值优化学习,对其前件及后件的结构和学习算法进行改进,使T-S-FNN的计算简化、鲁棒性更强。利用改进的T-S-FNN对MMW图像进行处理,实验结果表明,该模型具有较好的非线性噪声抑制能力。To denoise efficiently nonlinear noise existing in milli-meter wave（MMW）images,the property of the T-S fuzzy neural network（T-S-FNN）model that it can efficiently distinguish uncertain information was utilized to implement the approximation of noise existing in nonlinear information of MMW images to achieve the goal of denoising.To avoid the defect of redundant fuzzy rule of common T-S-FNN model,in the premise part network,considering the constraint of membership function based on the self-adaptive fuzzy cluster,and in the consequent part network,considering the weight optimization,the structure and training method of the premise part network and the consequent part network were modified.The modified T-S-FNN model has simpler calculation and better robustness.Utilizing the modified T-S-FNN model to process MMW images,experimental results show that the modified T-S-FNN model can well depress noise existed in non-linear information.

关 键 词：非线性信息 模糊神经网络 T-S模糊模型 毫米波图像 图像消噪 

分 类 号：TN911.73[电子电信—通信与信息系统]