针对低信噪比下红外小目标的特点,提出一种改进的实用动态规划算法,通过适当的预处理抑制背景噪声、多帧图像灰度累加以提高信噪比,并由虚警和检测概率确定分割门限,检测出小目标位置及状态.详细地叙述动态规划的计算过程及分割门限的确定方法,以简单的步骤实现对低信噪比小目标的检测.实践证明,改进的动态规划算法不仅能完成检测的任务,而且并行性好,易于硬件实现,为下一步的目标跟踪打下了良好的基础.

An update dynamic programming algorithm is presented in allusion to the features of small moving object.Using the sum of grey level in image sequences, the target is brought into prominence and noise is suppressed by the image preprocessing at the same time. The probability of detecting and false alarm are used to determine the threshold of image, which can abstract the small target accurately. And dynamic programming algorithm performs well in detecting, and it lays foundations for the tracking in the next step.