基于自适应空间滤波多级假设检验的空间目标识别方法  

Space Target Recognition Method Based on Adaptive Spatial Filtering Multistage Hypothesis Testing

作　　者：苗澍茁 范存波[1] 温冠宇[1] 高健[1] 赵国海 MIAO Shuzhuo;FAN Cunbo;WEN Guanyu;GAO Jian;ZHAO Guohai(Changchun Observatory/NAO,Chinese Academy of Sciences,Changchun 130117,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院国家天文台长春人造卫星观测站,长春130117 [2]中国科学院大学,北京100049

出　　处：《光子学报》2021年第11期359-368,共10页Acta Photonica Sinica

基　　金：国家自然科学基金委与中国科学院天文联合基金项目(No.U1731240);国家自然科学基金委面上项目(No.11973064);吉林省与中国科学院科技合作高新技术产业化专项资金项目(No.2020SYHZ0049)。

摘　　要：针对目前空间目标检测技术存在目标识别难度大、虚警率高、算法效率低等问题,提出了一种基于自适应空间滤波多级假设检验(ASMHT)算法的空间目标识别方法,用于提取光电观测系统在恒星跟踪模式下的空间目标,包括目标粗筛选与目标精筛选两过程。目标粗筛选使用尺度空间高斯差函数得到预处理后星图中各候选目标的尺度值,并将目标尺度值作为空间滤波窗口的尺度大小,利用连续多帧空间滤波窗口内不同种类目标的灰度分布特征代替灰度相关准则,去除无像移的背景恒星和随机噪声,筛选出疑似运动目标。目标精筛选使用改进的多级假设检验方法,通过建立疑似运动目标速度搜索窗极大提高了算法效率,最后依据轨迹特征筛选出空间目标。空间目标仿真星图结果表明,与现有空间目标识别方法相比,ASMHT算法具备空间目标低信噪比条件下的目标识别能力,且综合检测性能最好,在相同的虚警率下可获得更高的检测率。真实星图测试比较结果表明,ASMHT算法计算复杂更低,实现了更加精确的空间目标识别。Considering that the current space target detection technology has some problems in target recognition,which are high false alarm rate and low algorithm efficiency,a space target recognition method based on Adaptive Spatial Filtering Multistage Hypothesis Testing(ASMHT)algorithm was proposed,which is used to extract space targets from star image in the sidereal tracking mode in photoelectric observation system.The method includes two processes of target coarse screening and target fine screening.Target coarse screening uses Gaussian difference function based on the scale space to get the scale value of each candidate target in the preprocessed image.The scale value is also set to the size of the spatial filtering window,and the gray correlation criterion is replaced by the gray distribution characteristics as the accordance of different kinds of targets in the spatial filtering window.The background stars without image motion and random noise are removed,and the suspected moving targets are screened out.The improved multistage hypothesis testing method is used in target fine screening,in order to greatly improve the efficiency of the algorithm,the speed window of candidate targets are established and the space targets are selected according to the trajectory characteristics.The experimental results of space target in simulation image show that,compared with the existing space target recognition methods,the ASMHT algorithm has the ability of target detection under the condition of space targets in low SNR,and the comprehensive detection performance is the best,which can also obtain higher detection rate under the same false alarm rate.Real star image test results show that the ASMHT algorithm has lower computational complexity and achieves more accurate space target recognition.

关 键 词：空间目标检测 目标识别 空间滤波 尺度空间理论 多级假设检验 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]