受多种因素的影响, 白天获得的红外星图像信噪比低, 且背景通常是不均匀的, 为红外星目标的提取造成了极大的困难。通过理论计算, 实际白天拍摄的近红外星图信噪比极低, 使用一般的滤波方法无法提取出恒星目标, 目前常用的形态学方法对于星图像的处理也不甚理想。首先分析背景特性, 采用多帧叠加的方法削弱随机噪声, 增大信噪比; 然后设定阈值对叠加后的星图进行背景消除, 得到只含有目标及噪声的图像; 最后基于图像的特性, 使用改进的 SUSAN算子对星图进行目标检测, 进而分割出目标, 实验证明, 与传统方法相比, 该方法可较好地分离出恒星目标。

Affected by many factors, the infrared star image has a low SNR, and the background is nonuniform, which puts great difficulties in the process of picking up star target. According to theoretical calculation, actual star image has too low SNR to use the traditional methods to process it. Currently used morphology algorithm for the processing of star imagery is not good enough. The paper first analyzes the background characteristics, using multi-frame superimposition to weaken the random noise thus increasing the signal-to-noise ratio, and then sets the threshold to eliminate background of superimposed star maps to obtain the image only containing the target and noise. Finally, based on the characteristics of the image, we used the improved SUSAN operator for target detection and then extract the target. The experiment proved that the method can better isolated star targets compared with traditional ones.