空间目标由于距地相对较远, 且散射光信号受到大气介质的强散射, 在地基测量中很难获取到目标的准确信息。 近年来光谱观测技术蓬勃发展, 由此为空间目标测量提供了新的方案, 但在采集的目标光谱信息中, 由于目标轨道高度、 材料组成等大多相近, 很难直接从光谱曲线中分辨出目标。 为此基于双向反射分布函数（BRDF）散射理论, 建立了空间目标散射光谱成像模型, 并由1.2 m口径地基观测平台与光谱视频成像系统实验测量了一组高轨道同步卫星（GEO）目标, 光谱范围为400～720 nm, 光谱分辨率为2 nm。 采用径向基神经网络算法对光谱数据中的BRDF进行解混, 实验测量了六种空间目标典型材料的BRDF。 由于目标相对较远, 已经超出探测系统的衍射极限, 因此目标可视为点目标, 但在地基测量中大气层是阻隔在探测系统和目标之间的重要屏障, 目标光信号穿过大气层时会受到大气介质的强烈散射, 这种散射虽然很大程度上削弱了光信号, 但同时光信号也被按原结构放大。 依据光学记忆效应, 目标光信号穿过均匀大气介质后其结构仍保持不变。 基于以上分析, 目标光斑图像应该保留有目标投影结构的信息。 为此采用针对目标光斑图像纹理区域分割反演的方法, 将目标光斑划分为10个纹理区域, 并提取对应光谱数据。 通过探测系统传递函数标定以及减噪处理, 获得了观测时段在轨目标空间几何角度下的光谱曲线。 再利用建立的典型材料光谱数据库进行拟合反演。 结果表明: 在2号、 5号、 10号纹理区域反演出了区别于其他区域不同的材料类型。 同时, 反演的各纹理区的材料面积比也有较大不同。 为进一步评估拟合结果, 采用非奇异矩阵对拟合效果进行评价, 分析了扰动方程, 拟合准确率最高为85.283 3, 最低为76.982 7。 这说明拟合结果是相对真实的, 目标散斑图像中含有可分辨的目标投影结构信息。 此研究为揭开点目标成像探测和散斑图像结构识别提供了新的方向。

Because the space target is far away from the ground and the atmospheric medium strongly scatters the scattered light signal, it is difficult to get the accurate information of the target in ground-based measurement. In recent years, the rapid development of spectral observation technology has provided a new method for measuring space targets. However, it is difficult to distinguish the target directly from the spectral curve in the collected target spectral information because the target orbital height and material composition are mostly similar. Therefore, based on the bidirectional reflection distribution function (BRDF) scattering theory, the scattering spectral imaging model of space target is established. A group of geosynchronous(GEO) targets were experimentally measured by a 1.2 m aperture ground-based observation platform and spectral video imaging system. The spectral range is 400 to 720 nm, and the spectral resolution is 2 nm. A radial basis neural network algorithm is used to unmix the BRDF in spectral data. The BRDF of six typical materials for space targets is measured experimentally. Because the target is relatively far away, it has exceeded the diffraction limit of the detection system so that the target can be regarded as a point target. However, in ground-based measurements, the atmosphere is an important barrier between the detection system and the target. The target light signal will be strongly scattered by the atmospheric medium when passing through the atmosphere. This scattering greatly attenuates the light signal, but at simultaneously the light signal is amplified according to its original structure. According to the optical memory effect, the structure of the target optical signal remains unchanged after passing through the uniform atmospheric medium. Based on the above analysis, the target spot image in measurement should retain the information of the target projection structure. Therefore, a method of segmentation inversion for the texture region of the target light spot image is used to divide the target light spot into 10 texture regions and extract the corresponding spectral data. Through the transfer function calibration and noise reduction processing of the detection system, the spectral curve of the space geometry angle of the orbiting target in the observation period is obtained. Then the typical material spectral database is used for fitting inversion. The results show that the material types in texture areas No.2, No.5 and No.10 are different from other areas. At the same time, the material area ratio of each texture area is different. To further evaluate the fitting results, a non-singular matrix was used to evaluate the fitting effect, and the disturbance equation was analyzed. The highest fitting accuracy was 85.283 3, and the lowest was 76.982 7. It shows that the fitting results are relatively real. Target speckle image contains distinguishable target projection structure information. This study provides a new direction for detecting point target imaging and speckle image structure recognition.