Abstract: A new technology on real-time detecting of infrared dim moving point targets in image sequences is studied.The direction prediction of trajectories is adopted which reduces calculation amount and calculation time greatly.The energies along the candidate trajectories are accumulated and enhance the SNR（Signal Noise Ratio） of the images.Angle restriction and inter-frame association is used after the constant false alarm rate detecting，and gets the possible targets more accurately.The experimental results show that the technology can detect a number of moving targets with arbitrary trajectories accurately in real-time and have good performance.

Key words: direction prediction, infrared image sequence, point target detection, real time

摘要： 在红外图像序列情况下，研究了一种直接在三维时空中实时检测弱点状目标的技术。采用轨迹的方向预测，减少了计算量，提高了实时性；采用目标航迹的能量累积，提高图像信噪比；在恒虚警率检测条件下，采用角度约束和帧间关联获得了可能的目标轨迹。实验结果表明：该技术能够准确检测出任意轨迹运动的多个点目标，具有良好的实时性和准确性。

关键词: 方向预测, 红外图像序列, 点目标检测, 实时性