一种天文光电图像序列弱小目标实时检测算法

A Real-time Algorithm for Weak Small Targets Detection in Astronomical  Opto-electronic Image Sequences

  • 摘要: 地基大视场天文光电观测系统获取的序列图像中,空间目标和数目众多的恒星目标成像特性相似,大部分都表现为弱、小目标特性,实现它们稳健实时检测具有较大的难度。本文在深入分析天文光电图像弱小目标成像特点的基础上,结合实际工程对目标检测的强实时性和高检测概率需求,针对恒星目标和空间目标的时空域特性差异,提出了一种基于时空域联合滤波的弱小目标检测算法,首先在空域上通过抑制背景增强目标,然后在时域上分别通过多帧相关运算和多帧差分运算,快速有效地检测恒星目标和空间目标。最后结合实测数据和硬件平台,验证了该方法的有效性及实时性。

     

    Abstract: In the image sequences obtained by the large field of view astronomical opto-electronic observation system on terrestrial surface, the image characteristics of the space targets and the large amount of stars are similar, and the targets mostly appear weak and small. It is difficult to realize their robust real-time detection. According to the detailed analysis of the characteristics of the weak small targets in the astronomical opto-electronic images, the article proposes a weak small targets detection algorithm of temporal-spatial combined filtering, aiming at the time-space domain characteristics of the stars and space targets and combining the stern real project’s need of real-time performance and detection of probability. The algorithm suppresses background and highlights targets in spatial, and detects the stars and space targets respectively by multi-frame correlation and multi-frame difference concerning the different moving characteristic in temporal. Finally, the combination of detection based on the real-measured data and the realization on hardware platform shows the algorithm’s validity and real-time quality.

     

/

返回文章
返回