摘要
针对嫦娥三号所拍摄的图像为近月表图像,未搭载激光高度计,无法获得高程值这一问题,提出一种嫦娥三号高程值估计的方法。该算法是在嫦娥二号多传感器数据的基础上,训练出在图像中的特征描述子与高程值之间对应关系的误差反向传播神经网络模型;然后利用嫦娥三号的高精度图像特点,估计出相应的高程值。实验测试结果表明,这种算法可将高程值估计误差降低到3.94%,因此得到的嫦娥三号高程值具有可靠性,可应用于高精度月球重建。
According to the image taken by the Chang’e-3 as a near moon image,the elevation value cannot be obtained without a laser altimeter,and a method to estimate the elevation value of the Chang’e-3 is proposed.The method is based on the"Chang’e-2"multi-sensor data and trains the BP neural network model of the corresponding relationship between the feature descriptors and the elevation values in the image.Then,the corresponding elevation values are estimated using the features of the high precision image of the Chang’e-3.Exxperimental results show that the proposed method can reduce the elevation value estimation error to 3.94%.Therefore,the elevation value of the"Chang’e-3"is reliable and can be applied to high-precision moon reconstruction.
作者
周哲哲
赵萌
石凡
陈胜勇
栾昊
ZHOU Zhezhe;ZHAO Meng;SHI Fan;CHEN Shengyong;LUAN Hao(School of Computer Science and Engineering, Tianjin Univ. of Technology, Tianjin 300384, China;Key Lab. of Computer Vision and System of Ministry of Education, Tianjin Univ. of Technology, Tianjin 300384, China)
出处
《西安电子科技大学学报》
EI
CAS
CSCD
北大核心
2019年第2期139-144,共6页
Journal of Xidian University
基金
国家自然科学基金(61703304
U1509207)
关键词
嫦娥三号
高程值估计
误差反向传播神经网络
全月重建
Chang’e-3
elevation value estimation
back propagation(BP) neural network
moon reconstruction
