从图像序列中检测定位小目标的一种方法

A Method for Detecting and Locating Small Target from Image Sequences

研究红外成像小目标检测问题．高通滤波器滤去低频图像背景信号，再在高频图像信号中使用似然比检测方法提取可能的目标点．考虑到目标运动的连续性和约束条件，将小目标的检测问题化为聚类问题，并通过设置置信度函数，搜索可能的目标点集以得到真正的目标点集，并检测出小目标的位置．然后使用最小二乘法，对目标位置进行校正．实验结果表明，该方法能有效地检测定位运动小目标，且具有很强的抗干扰性．

Detection scheme of small moving targets for infrared imaging is studied. Low frequency background signal can be suppressed by means of high pass filter. Possible target points can be abstracted form image signals of high frequency with likelihood ratio test method. With concern about continuity and constraint of target movement, small target detection problem can be reformulated as an optimal clustering one. By setting a possibility function, the real target points set can be attained and the position of small target can be detected as well. It is sequentially corrected with the use of least square method. The experiment results indicate that our method can detect and locate small moving target validly, showing strong robustness.