光谱学与光谱分析, 2019, 39 (7): 2013, 网络出版: 2019-07-23  

均匀背景中的动态小目标光谱数据获取方法研究

Spectrum Acquisition of Dynamic Small Objects in Uniform Background
作者单位
1 北京航空航天大学仪器科学与光电工程学院, 北京 100191
2 中国科学院光电研究院中国科学院计算光学成像技术重点实验室, 北京 100094
摘要
光谱是一种可以表征物质特性的光学信息, 利用光谱成像仪可以获取处于视场范围内的物质的光谱图像, 成熟的光谱成像技术均需要通过多次采集才能够获取完整的光谱图像数据立方体, 相应系统的时间分辨率比较低, 不适用于动态目标的光谱获取。 快照式光谱成像在动态目标光谱成像方面具有较大的优势, 其中编码孔径快照光谱成像技术是一种将压缩感知计算方法融入到光谱成像过程和图谱重构过程中的光谱成像技术, 在采样过程中完成数据压缩, 具有高通量优势, 可以利用单次曝光的混叠数据, 重构出目标光谱数据立方体, 实现快照式成像, 使得对动态的目标进行监测成为可能。 实现监测需要目标的信息满足稀疏性的假设, 实际目标很难满足这样的条件, 重构误差比较大, 不利于对动态的小目标进行监测和识别。 针对均匀背景中动态小目标的光谱数据获取, 提出一种双色散通道的编码孔径光谱成像方法, 系统由两个通道组成, 每个通道均包含一个光谱仪, 其色散方向互相垂直, 并共用一个前置望远镜系统和编码孔径。 该系统可以实时观测均匀背景区域中的动态小目标。 由于两个通道的色散方向互相垂直, 可以从背景中分离出小目标的位置和相对应的编码。 假设目标出现在视场中前后, 背景的辐射特性变化很小, 利用目标出现前的数据计算出背景光谱; 目标出现后, 通过帧间差分运算, 消除背景辐射的影响, 提取出目标位置对应色散区域中数据, 利用约束最小二乘算法, 重构运动小目标的光谱数据立方体。 进行光谱数据重构, 进行背景光谱补偿后, 获得完整的动态小目标光谱数据。 文章对成像过程建立了数学模型, 并对重构方法进行了仿真验证, 结合编码孔径的统计特征, 使目标随机出现在不同的位置, 统计重构光谱的峰值信噪比概率分布, 并调整目标尺寸, 分析目标尺寸对重构精度的影响, 最后与编码孔径成像系统的两步软阈值迭代算法重构结果进行了对比。 结果表明, 这种方法在均匀背景中, 采用随机编码矩阵进行编码, 目标尺寸小于5×5个像元时, 相对于编码孔径成像系统, 提高了目标的信息重构精度和概率, 并且极大的减小了运算量, 可以实现对运动目标的实时监测。
Abstract
Spectrum is a kind of optical information that represents the properties of the material. By using spectral imager, the spectral image of the objects can be obtained in the field of view. At present, mature spectral imaging techniques require multiple snapshot to get a complete spectral image data cube. The time resolution of the corresponding system is relatively low, which is not suitable for the spectral acquisition of dynamic target, and the acquisition of dynamic target spectrum requires a snapshot imaging technique. Snapshots of spectral imaging has great advantage in dynamic target spectral imaging, the coded aperture snapshot spectral imaging (CASSI) technology is one method that integrates a compressive perceptual computation methodology into the spectral imaging procedure and the data cube reconstruction process. In the process of imaging, the data compression is completed, and CASSI has the advantage of high throughput. It is possible to reconstruct the target spectral data cube and realize the snapshot imaging by using the single exposure data, which makes it possible to monitor the dynamic target. But the targets information is difficult to prove the sparse hypothesis of this method, leading to a large reconstruction error, which is unfavorable to the monitoring of dynamic small targets. On the basis of CASSI, a new method of double dispersion channel coding aperture spectral imaging system is presented, which is used to obtain the spectral data of dynamic small targets in uniform background. The system consists of two channels, each containing a spectrometer whose dispersion directions are perpendicular to each other, and shares a front-end telescope system and coded aperture. Because the dispersion directions of the two channels are perpendicular to each other, the position of the small target and the corresponding coding can be separated from the background for spectral data reconstruction. So this new system can observe small dynamic targets in uniform background area in real time. Assuming that the radiation characteristic of the background changes little before and after the target appears in the field of view, the background spectrum can be calculated using the data before the target appears. And the target spectrum can be recovered through specific algorithm by utilizing the separated data and background spectrum. An imaging process mathematical model is established, and the reconstruction method has been tested. The target was made to randomly appear in a different position, and the PSNR probability distribution of the reconstructed spectrum was speculated. After adjusting the target size, the influence of target size on the reconstruction accuracy was analyzed. At last, the results were compared with the results of TwIST method reconstruction of coded aperture imaging system. Results show that this method improves the precision of target information recovery accuracy, and greatly reduces the computational complexity to realize real-time monitoring for moving targets when target with less than 5×5 pixels in homogeneous background.

王建威, 赵剡, 李伟艳, 裴琳琳, 孙建颖, 孙成明, 吕群波, 刘扬阳. 均匀背景中的动态小目标光谱数据获取方法研究[J]. 光谱学与光谱分析, 2019, 39(7): 2013. WANG Jian-wei, ZHAO Yan, LI Wei-yan, PEI Lin-lin, SUN Jian-ying, SUN Cheng-ming, L Qun-bo, LIU Yang-yang. Spectrum Acquisition of Dynamic Small Objects in Uniform Background[J]. Spectroscopy and Spectral Analysis, 2019, 39(7): 2013.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!