基于时序多帧投影的空间目标检测算法

为了对星图中空间目标进行检测识别，对基于 SURF（Speed-Up Robust Featrues）算法的星图精确配准技术和美国 SBV（Space-Based Visual）计划采用的 MTI（Moving Target Indicati-or）在轨目标检测算法进行了深入研究，提出一种针对16 Bits 星图的多目标检测算法，具体包括：首先利用 SURF 算法提取序列星图的特征点，根据最小二乘法计算得到的全局运动参数对星图进行精确配准；然后利用一种改进的 MTI 算法对序列星图进行时序多帧投影以抑制背景，得到仅含有疑似目标的序列图像；最后经过目标初始运动状态的建立，速度滤波以及坐标插值得到目标的运动轨迹。利用实拍的20帧序列星图验证算法性能，经本文算法配准后，星像质心的均方误差（RMSE，Root Mean Square Error）最小达到0.3269 pixel，平均值为0.5441 pixel；序列图像中的3个运动目标均被检出，且无虚警。实验结果表明，本文配准算法的精度能够满足时序多帧投影的要求，且目标检测算法符合恒虚警原理。

Space target detection algorithm based on sequential multi frames projection

In order to detect and recognize the space targets in star map,precise registration algorithm based on the SURF and the MTI detection algorithm in American SBV program are studied. A multiple target detection algorithm aiming at 16 Bits star maps is proposed. Firstly,the matching feature points of image sequences are extracted with the SURF algorithm. The global motion parameters calculated with the least square method are used to precisely registrate the star maps. Then an improved MTI algorithm of time series multi frame projection is used to restrain the background. After projection,only latent targets are retained. Finally,after establishing the initial motion state,targets trajectory are obtained through velocity filtering and coordinate interpolation. The performance of algorithm is verified with 20 real star images. After registration,the minimum RMSE of stars centroid is 0.3269 pixel,and the average value of RMSE is 0.5441 pixel. 3 moving objects in image sequence are all detected,and no false alarm arises. The experimental results show that,the registration accuracy can meet the requirement of multi frame projection,and the detection algorithm confirms to constant false alarm rate principle.