基于动态规划的空间小目标检测算法

以往对动态规划法检测小目标算法的研究存在对中途入场和出场目标检测能力弱的问题,而且多数停留在对仿真图像的实验上,与实际应用还有一定差距。将动态规划算法应用于空间小目标检测,并从三方面对其进行改进:第一,针对工程应用中人为增加光学散焦导致目标信息丢失的问题,对递归方程进行修改,提高目标能量累积效率;第二,将各个速度平面分别计算修改为速度更新,减少计算量;第三,增加对入场和出场目标的处理,克服不能有效检测中途入场和出场目标的问题。实摄序列图像对比实验显示,该方法能够提高动态规划算法的检测效率,有效检测入场小目标。

Space Small Targets Detection Based on Dynamic Programming Algorithm

The former studies on small targets detection based on dynamic programming algorithm couldn't detect the entrance and exit targets efficiently. Most of them still experimented with simulation images and had distance to real application. In this paper, Tonissen's dynamic programming algorithm was applied in space small targets detection and was improved on the following three aspects focused on the problems. First, focus on the case of defocus which usually appeared in the project and result in information lost, the recursive equation was improved to rise the efficiency of energy accumulation. Second, the multivelocity plane calculation was simplified as velocity refresh to reduce the computational complexity. Third special process was added to solve the problem of entrance and exit targets. After experiment with real space sequence images, the result shows that the algorithm was efficient to detection and especially to the entrance and exit targets.