焦平面阵列BP神经网络非均匀性校正及其算法改进

BP Neural Network Nonuniformity Correction Technique for FPA and Improved Algorithm

在紫外成像系统中,紫外焦平面阵列的非均匀性是影响成像质量的重要因素。介绍了标准的神经网络算法对焦平面阵列非均匀性的校正,并针对标准的神经网络算法的收敛速度慢的缺点,提出了改进算法。通过matlab对算法进行仿真,结果表明BP神经网络(Back-Propagation Neural Network)算法对焦平面阵列的非均匀性有良好的校正效果,改进后的算法效率有了较大地提高。神经网络非均匀性校正算法可以广泛的运用于其他焦平面阵列的非均匀性的校正中。

In UV image systems,the nonuniformity between individual detectors in UV focal plane arrays is the important problem,which affects the quality of image.This paper presents the standard neural network nonuniformity correction algorithm and the improved algorithm according to the flaw of the standard neural network nonuniformity correction algorithm.The algorithms were simulated in matlab.The result showed that the standard neural network algorithm can correct the nonuniformity effectively and the improved algorithm can advance the efficiency high.Back propagation neural network nonuniformity correction technique can widely be applied to the nonuniformity of some other FPA.