节点文献
多光谱、超光谱成像探测关键技术研究
Research of Key Technologies of Multispectral, Hyperspectral Imaging Detection
【作者】 许洪;
【导师】 王向军;
【作者基本信息】 天津大学 , 测试计量技术及仪器, 2009, 博士
【摘要】 多光谱、超光谱成像探测技术是新一代光电探测技术,该技术利用具有一定光谱分辨率的多光谱、超光谱图像进行目标探测,在目标材质识别、异常目标检测、伪装目标辨识、复杂背景抑制等目标探测技术领域都有着极为重要的应用。小型化、实用化的成像光谱仪的研制和相关探测性能、探测方法的研究是成像光谱技术研究领域的一个主要趋势。本文正是以此为核心,深入进行了如下方面内容的研究:1.综合考虑目标红外辐射特性、大气传输透过率、色散式光谱仪辐射传递特性、选用红外探测器参数、数据链传输等环节,研究了成像光谱探测系统的辐射能量传递模型。研究了成像光谱探测器光谱分辨率和作用距离的表达公式。选用多种典型地物红外发射率数据,非制冷微测辐射热计红外焦平面阵列参数,利用大气辐射传输计算软件生成大气红外透过率曲线,根据所建立模型,计算了以非制冷微测辐射热计阵列为探测器的色散式分光的成像光谱仪的光谱分辨率参数。2.研究了一种基于可见光/近红外波段多光谱、超光谱图像数据的地面场景建模方法,探讨了无监督分类方法和有监督分类方法相结合的像元地物分类、匹配、标记的策略,可以高效地解决像元地物自动匹配标记的问题。利用RGB彩色图像验证了这一方法,在将图像分割后为每类像元赋予相应的红外发射率数值,生成了四个红外波段的多光谱仿真图像,验证了该方法的可行性,探讨了多光谱、超光谱图像数据在仿真应用中的各自特点。在仿真结果中,观察到了不同波段图像中目标和背景之间呈现不同特征,利用SCR指标对目标探测性能进行了评估,结果表明在窄带特征波段可以有效提升探测性能。3.基于共像面分割的光学原理,利用光学仿真技术设计构建了一种基于楔形转像棱镜实现共像面分割的小型化四波段光谱成像实验光路装置。利用构建的四波段光谱成像光路装置进行了颜色相近目标的辨识实验研究,试验中通过窄带滤光片在不同光谱波段成像,利用聚类技术对目标的多光谱图像数据进行了分析,结果表明采用目标的多光谱图像数据可以可靠地区分颜色相近的不同目标,在真伪目标识别应用中具有一定意义。
【Abstract】 Multispectral/hyperspectral imaging detection technology is one of the new generation of photoelectric detection technologies which can use multispectral/hyperspectral image data with different levels of spectral resolution to discriminate targets. It has significant applications in many fields of detection technologies such as material identification, anormaly detection, false-target discrimination, backgrounds suppression.Development of small size, practical imaging spectrometers and research of their detection abilities and relevant detection methods are the leading trends in the future. These mentioned technologies are the key research contents in this paper, which are:1. A integrated radiative transfer model is researched considering characteristic of target infrared radiation, the effect of atmospheric transmission, characteristic of energy transmission of dispersive spectrometer, performance parameters of the infrared detector array, as well as data communication. The parameters of spectral resolution and detection range of the imaging spectrometer are analyzed according to the radiative transfer model. Selecting infrared emissivity data of several typical ground objects, the performance parameters of the state-of-the art uncooled microbolometer FPA, utilizing LOWTRAN code to generate the atmospheric transmittance, the value of spectral resolution of a LWIR dispersive imaging spectrometer is calculated based on an uncooled microbolometer FPA.2. A ground scene modeling method is researched based on multispectral or hyperspectral image data in VIS/NIR band. A combined strategy of unsupervised and supervised is put forward to efficiently solve the auto-matching and marking of pixels. The method is verified by RGB images. Simulation images in four LWIR bands are generated after pixels segmentation and infrared emissivity assignment. The respective characteristis of multispectral and hyperspectral image data in pixel classification and matching are pointed out. In simulation results, the difference between targets and backgrounds are observed. Using the index of SCR to evaluate the detection performance, it is shown that the detection performance can be effectively improved in certain narrow spectral band.3. A small size, four bands multispectral imaging experimental optical facility is constructed based on the optical technique of imaging surface division, taking advantage of the tools of optical simulation. Using the designed multispectral imaging experimental optical facility, the research experiments of discrimination of targets with similar colors are carried out. Using the clustering technology to analyze multispectral image data, the result shows that color similar targets can be reliably discriminated by multispectral data.
【Key words】 multispectral; hyperspectral; uncooled microbolometer; multispectral image simulation; imaging surface division;