基于模式侧抑制复杂背景下的小目标检测

Small target detection in clutter based on pattern lateral inhibition

在复杂背景的图像序列中检测小目标一直是研究重点。提出了一种基于模式侧抑制的复杂背景下小目标检测的方法。将复杂背景下的小目标检测问题看作是在大量相似模式中寻找某个特殊模式的模式识别问题。通过建立基于模式的侧抑制网络模型,在模式空间中对模式进行侧抑制处理,达到突出特殊模式抑制相似模式的目的。同时指出,用该方法对图像进行多级处理可进一步改善处理效果。进行了该方法与去局部均值滤波、基于形态学Top!Hat算子滤波、多级滤波和基于侧抑制理论滤波四种传统方法的比较实验,并将信杂比(SCR)及信杂比增益作为评价算法性能的指标。实验结果表明,提出的方法在提高图像信杂比方面要优于其他几种方法,能有效抑制背景杂波,提高对目标的单帧检测能力。

It is an important task to detect the small targets in a sequence of images against complex background. An algorithm of small target detection based on pattern lateral inhibition is proposed.Small target detection in complex background is considered as a problem of finding the special patterns in similar patterns. The special patterns are enhanced with the Lateral Inhibition network. At the same time,multi-level processing is presented to improve treatment effects.The performances of several traditional filters and the method in this paper are compared.Two defininitions of SCR and SCR gain are defined to evaluate the performance of the approach Experimental results show that the approach proposed is effective in detecting the small targets in clutter.