基于神经网络的航天光学遥感器在轨信噪比测试方法

提出一种基于神经网络的航天光学遥感器在轨信噪比的的测试方法。通过模拟得到了大量的包含有不同信噪比等级的遥感图像,并将其作为网络训练和测试的样本。通过对遥感图像进行分析,找到了分别与景物结构和噪声有关的特征向量,并将其作为神经网络的输入。在对大量样本图片进行训练后,可完成对由遥感器传输下来的任意一幅地面景物图像进行信噪比的测试,从而避免了传统方法对特定地面景物目标在成像测量中的诸多弊端,平均测量误差约为10%。

Assessment of signal to noise ratio of space optical remote sensor based on neural network

A method of assessing the signal to noise ratio (SNR) of space optical remote sensor (SORS) based on artificial neural network (ANN) is presented.A large amount of remote images with different SNR level are simulated,they are used for training and testing of ANN.Through analyzing the image of remote sensor,eigenvectors related to the structure of landscape and noise are found out respectively.After training,the ANN could assess the SNR through images of any landscape,thereby the defects are avoided in the traditional method which needs imaging special views.The mean assessment error is about 10%.