基于光谱压缩的线阵推帚式超光谱成像机理研究

Hyperspectral imagers use a 2-dimensional detector domestically and abroad.Spectral data is acquired in across-orbit direction,and spatial information is acquired in perpendicular to the moving satellite.Because some developed countries make a high technology blockage to us, we are difficult to get 2-dimensional detectors with big dimension,high frame rate at the present time.For the next decade, we can hardly get or manufacture 2-dimensional detectors with high capability.That restricts application of hyperspectral imagers in space exceedingly internally.. This project follows closely internationally advanced developed direction, .find another excellent approach to new method of computational spectral imaging.Bringing forward research on mechanism of hyperspectral imagers with linear detector imager system on spectrum compressed. Compressing the quondam spectral information to one row after signal sparse representation.By the method of characteristics decoupling, making high precision reconstruction of compressed data.Using low dimension projection of the objects to reconstruct high dimension numeral model of the objects. Making use of an 1-dimension detector instead of a 2-dimension detector to get a 3-dimensional data cube holding the spectral information and the spatial information of each ground object.That reduces the dimension of the detectors and complexity of imaging system greatly. There is important and far-reaching theoretical significance and extensive applied value to .improve the technique level of digital remote sensing,reduce depending on the overseas technology seriously,satisfy exigent requirement for remote sensing hyperspectral images internally.

超光谱成像仪目前在国内外均采用面阵探测器，在沿轨方向获取目标的光谱信息，在穿轨方向获取目标的空间信息。由于某些发达国家对我国实行高技术封锁，目前我国在大规模、高帧频面阵探测器方面获取困难，今后的十多年我国几乎无法获得或制造出高性能的面阵探测器，极大程度限制了我国超光谱成像技术的空间应用。.本项目紧跟国际前沿发展趋势，另辟蹊径，探索新型的计算光谱成像方法，提出基于光谱压缩的线阵成像体制超光谱成像机理研究，即将原来的光谱维信息稀疏采样后压缩成一列，再通过特征解耦方法对压缩数据进行高精度重构，实现利用目标的低维投影，重建目标的高维数字模型，采用线阵探测器替代面阵探测器获取目标的空间和光谱数据立方体，大大降低了探测器规模及成像系统的复杂度。这对于提高国内数字遥感技术的水平，降低对国外同类产品的严重技术依赖，满足国内对超光谱遥感图像的迫切需求，具有极其重要而深远的理论意义和广泛的应用价值。

本项目针对国内大规模、高帧频面阵探测器发展的制约，以压缩感知理论为基础，突破传统光谱成像模式，研究了实现超光谱成像与压缩感知融合的光谱压缩线阵推帚式超光谱成像理论与方法，从而大大降低探测器规模及成像系统的复杂度，使线阵探测器能够获取超光谱成像的三维数据立方体，改变以往必须采用面阵探测器实现超光谱成像的局面。本项目通过灵活的编码孔径成像系统设计，有效改善重构光谱图像质量，获取更高的峰值信噪比和更好的感知效果；提出联合结构稀疏表示和非负矩阵分解技术，充分利用高光谱图像谱间空间的冗余信息，得到更好的重建效果。针对现有多光谱图像重构技术中重构效果不理想、重构速度慢的问题，提出了一种基于双树复小波变换的多光谱图像重构方法。此外，本项目还提出了基于压缩感知的双通道遥感光谱成像和视频成像方法、基于互补编码模板的双通道编码感知光谱视频获取方法、基于压缩感知的天体光谱图像成像方法。