基于成像差分吸收光谱技术和压缩感知理论的烟囱烟羽断层重建研究

烟羽断层重建质量受两方面条件限制：其中一个限制条件是遥感设备的时间分辨率。以往的研究多使用多轴差分吸收光谱仪（MAX-DOAS）进行CT重建，受采集数据速度的限制，重建图像的时间分辨率较低。另一个限制条件是，采集到的数据量有限，是典型的不完全角度重建。过去多使用代数迭代重建算法或统计迭代重建算法，重建图像受测量误差的影响比较大，分辨率较低且伪影较多。构造了基于成像差分吸收光谱技术（IDOAS）的光谱数据采集系统，与多轴差分吸收光谱仪构造的系统相比，数据采集的时间分辨率提高了160多倍，基本解决了时间分辨率的问题。提出了一种基于压缩感知理论和低三阶导数模型的烟羽断层重建算法--投影凸函数集低三阶导数法，简称为POCS-LTD。在投影的过程中，使用代数重建算法使重建图像符合投影方程；在全变分迭代的过程中使用了优化算法，将低三阶导数模型的全变分归一化值作为优化算法的迭代方向，前次迭代运算结果与本次投影运算的差值的模作为迭代步长。对重建算法进行了数值模拟，并以重建图像的接近度和一致性相关因子为指标，对重建结果进行了分析。数值模拟表明，算法具有良好的抗误差能力，与传统的低三阶导数法相比，本文提出的算法将重建接近度减小了80%以上。使用烟羽数据采集系统进行了外场实验，用POCS-LTD算法对外场实验的数据进行了烟羽重建，重建图像显示烟羽图像清晰，伪影得到了较好的抑制。介绍的烟羽断层数据采集系统和烟羽断层重建算法，提高了烟羽断层重建图像的时间分辨率，减少了重建图像的伪影，扩大了光谱测量技术的应用范围。

Reconstruction of Stack Plume Based on Imaging Differential Absorption Spectroscopy and Compressed Sensing

The quality of computed tomography image of stack plume has two limits. Oneof the limitations isthe temporal resolution of data acquisition of remote sensing instrument. The conventional remote sensing instrument used for tomography is an multi-axis differential optical absorption spectrometer. Limited by its’ speed of data acquisition, the temporal resolution of the reconstructed image is low. The other limitation is the insufficient data acquired from remote sensing instruments. Algebraic reconstruction algorithms and iterative statistical algorithms are usually used in the reconstruction, but the reconstructed images have a low resolution and plenty of artifacts. To overcome the first limitation, data acquisition system in this paper is composed of imaging differential optical absorption spectrum technique. Compared with the system composed of the multi-axis differential optical absorption spectrum, the system’s temporal resolution increases above 160 times. An algorithm based on compressed sensing and a low third derivative model is introduced to overcome the second limitation. The algorithm is called projection on convex sets-low the third derivative method, which is POCS-LTD in short. The proposed algorithm belongs to gradient projection for sparse reconstruction algorithms, which is divide into two steps: projection and total variation iteration. In the process of projection, the algebraic reconstruction algorithm is used to make the reconstructed image conform to the projection equation, and the optimization algorithm is used in the process of total variation iteration. In the process of the total variation, the normalized value of the low third order derivative model is used as the iterative direction of the optimization algorithm, and the module of the difference between the result of the previous iteration and the present projection operation is used as the iterative step. According to nearness, the reconstructed images are evaluated, the relative difference of maximum and concordance correlation factor by numerical simulation. The numerical simulation shows that the proposed algorithm has good error resist,and reduces nearness by 80% compared with the traditional low third derivative method. The method is used to reconstruct the plume by the data gets from the field campaign. A clear plume and suppression of artifacts can be seen from the reconstructed image. The data acquisition system and algorithm introduced in this paper promote temporal resolution and reduce artifacts of the reconstructed images, and improve the technique’s practicability.