基于改进神经网络的IRFPA非均匀性校正方法

Non-Uniformity Correction Method of IRFPA Based on Improved BPNN

针对复杂背景下经典非均匀性校正算法不理想的问题,提出一种改进神经网络隐含层的非均匀性校正方法。该方法考虑邻域像素对中心像素的影响程度不同,利用模糊逻辑中的隶属度函数分类确定邻域像素的权值,对邻域像素分类加权处理作为期望输出。仿真试验结果表明,与几种经典校正方法相比,提出的方法在校正效果上有较大的提升。

We proposed an improved method of non-uniformity correction(NUC)of infrared focal plane arrays(IRFPA)on hidden-layer of Neural Network,to overcome the undesirable results in the complex background.Considering the different influence of neighboring pixels on the center pixel,the membership function of Fuzzy Logic was used to obtain the weights of neighboring pixels.Simulation results show that,the improved method have better correction results compare with classical methods.