| 引用本文: | 甘凤娇,罗成高,彭龙,等.利用卷积神经网络的太赫兹孔径编码增强成像.[J].国防科技大学学报,2022,44(1):14-21.[点击复制] |
| GAN Fengjiao,LUO Chenggao,PENG Long,et al.Enhanced terahertz coded aperture imaging using convolutional neural networks[J].Journal of National University of Defense Technology,2022,44(1):14-21[点击复制] |
|
| |
|
|
| 本文已被：浏览 6833次 下载 4811次 |
| 利用卷积神经网络的太赫兹孔径编码增强成像 |
| 甘凤娇1,罗成高1,彭龙2,梁传英1,王宏强1 |
|
(1. 国防科技大学 电子科学学院, 湖南 长沙 410073;2. 中国人民解放军78118.部队, 四川 成都 610000)
|
| 摘要: |
| 针对现有太赫兹孔径编码成像算法鲁棒性较差、计算求解复杂度较高等问题,提出基于卷积神经网络的太赫兹孔径编码增强成像方法。该方法通过构建一个端到端的神经网络来实现成像系统的隐式建模,利用网络强大的求逆能力和抗噪性能来实现低信噪比下的目标重构。通过仿真实验可以看出,该方法可以在不同信噪比下实现对不同稀疏度目标的重构。另外,与经典的优化迭代算法相比,该方法在低信噪比下可以实现对目标更高分辨率的重构。 |
| 关键词: 太赫兹 孔径编码成像 卷积神经网络 低信噪比 高分辨率 |
| DOI：10.11887/j.cn.202201003 |
| 投稿日期：2021-02-20 |
| 基金项目:国家自然科学基金资助项目(61701513,61971427,61871386,61921001) |
|
| Enhanced terahertz coded aperture imaging using convolutional neural networks |
| GAN Fengjiao1, LUO Chenggao1, PENG Long2, LIANG Chuanying1, WANG Hongqiang1 |
|
(1.College of Electronic Science and Technology, National University of Defense Technology, Changsha 410073, China;2. The PLA Unit 78118, Chengdu 610000, China)
|
| Abstract: |
| To address the problems of poor robustness and high computational complexity of some existing coded aperture imaging algorithms, an enhanced terahertz coded aperture imaging method based on convolutional neural networks was proposed in this paper. The method realized implicit modeling of the imaging system by constructing an end-to-end neural network that exploited the network′s strong inversion ability and noise immunity for target reconstruction at low SNRs (signal-to-noise ratios). The simulation experiments show that the proposed method can achieve the reconstruction of targets with different sparseness at different SNRs. Moreover, the proposed method can achieve a higher resolution reconstruction of the target at low SNRs in comparison with to the classical optimization iterative algorithm. |
| Keywords: terahertz coded-aperture imaging convolutional neural networks low signal-to-noise ratio high resolution |
|
|
|
|
|
