光谱编码计算关联成像技术研究

现有的多光谱成像技术通常采用光学分光的方式，使用多个探测器对成像场景的光谱图像进行采集，导致现有成像系统复杂，数据量大、效率低。针对现有技术的不足，提出基于正交调制模式的光谱编码计算关联成像技术。通过正交光谱编码矩阵融合Hadamard基图案构造投影散斑对宽带光源进行调制，单像素探测器收集成像物体与调制光源相互作用后的反射信号；应用演化压缩技术复原成像物体的混叠光谱图像；利用编码矩阵的正交性质解码出欠采样的光谱分量图像，对分离出的图像应用组稀疏压缩感知算法重构全采样的光谱分量图像，最后融合出成像物体的多光谱图像。通过数值模拟与实验两方面验证了所提方法的高效性。所提的技术简化了多光谱关联成像系统，降低了数据量。光谱编码方法可以扩展到更多的光谱通道，也可以应用在偏振关联成像、信息加密等领域。

Study on spectral encoded computational ghost imaging

The existing multispectral imaging technologies usually utilize optical spectroscopy and multiple detectors to capture spectral images. These techniques suffer from complexity, a large amount of data and low efficiency. Addressing these deficiencies, in this paper, a spectral encoded computational ghost imaging technology based on orthogonal modulation model was proposed. The orthogonal spectral encoded matrices fused with Hadamard patterns were used to produce the illumination patterns that modulate the broadband light source. A single-pixel detector was utilized to collect the back-reflected signal from the imaging objects. The evolutionary compressive technology was applied to recover the mixed spectral image. The subsampled spectral channel images were obtained from the mixed spectral image by means of the orthogonality of the spectral encoded matrices. Then the group sparse compressed sensing algorithm was applied to reconstruct the full-sampling spectral channel images, which finally fused the multispectral image of the imaging object. The efficiency of the proposed method was verified by a numerical simulation and an experiment. The proposed technology simplifies the multispectral imaging configuration and greatly reduces the amount of data. The orthogonal spectral encoded strategy can extend to more spectral channels and also can be applied to polarization imaging, information encryption, and other many fields.