摘要
由于复杂天气情况和夜间可视度低等场景限制,红外图像技术应用广泛,但是获取的成本较高。为降低红外图像的获取成本,利用二维傅里叶单像素成像的方法获得低采样率的红外图像,再利用条件对抗生成式网络进一步重构出清晰的红外图像。实验结果表明,在采样率为6.25%的情况下,可以重构出接近原始清晰图像的红外图像,因而利用关联成像降低红外图像获取成本的方法,具有较高的实用价值。
Due to complex weather conditions and low visibility at night,infrared image technology is widely used,but the cost of acquisition is high.In order to reduce the acquisition cost of infrared images,the two-dimensional Fourier single-pixel imaging method is used to obtain infrared images with low sampling rate,and then the conditional adversarial generative network is used to further reconstruct the clear infrared images.According to multiple experiments,the results show that at a sampling rate of 6.25%,infrared images close to the original clear images can be reconstructed,so the method of reducing the cost of infrared image acquisition by using associative imaging has high practical value.
作者
郭亮
GUO Liang(Security Department,Shanghai University of Engineering Science,Shanghai 201620,China)
出处
《智能计算机与应用》
2024年第5期144-149,共6页
Intelligent Computer and Applications
基金
国家自然科学基金(62275153,62005165)
上海市产业协同创新项目(HCXBCY-2022-006)。
关键词
关联成像
红外图像
深度学习
高分辨率重构
correlated imaging
infrared images
deep learning
high-resolution reconstruction