基于方向梯度的红外小目标检测算法

An Infrared Small Target Detection Algorithm Based on Derivative Gradient

飞行器拍摄的传统红外小目标检测方法实现简单,多适用于简单背景对于复杂背景,存在一定数量的虚警,难以甄别真假小目标。提出了一种基于方向梯度的红外小目标检测算法,融合了红外小目标的梯度各向同性、不光滑性和对比特性等3个主要特征,主要通过卷积操作完成,并对融合图像中的非0位置计算最小局部对比度,进一步增强小目标并抑制虚假目标。设计了4种背景下的红外小目标检测实验,同其他5种常用检测算法进行对比,实验结果表明,提出的检测算法虚警率较低,鲁棒性也较好。

The traditional infrared small target detection algorithm is easy to be accomplished,so they are more suitable for simple background;while under the complex background,there is a certain number of false alarm,and it is difficult to identify the true target from the false ones.In this paper,an infrared small target detection algorithm based on the derivative gradient is proposed.This algorithm combines such three main features of infrared small target as gradient isotropy,non-smoothness and contrastive characteristic of the infrared small target,and the combination is mainly implemented through convolution operation.The minimum local contrast of non-zero position in the fusion image is calculated to further enhance the small targets and suppress the false ones.In this paper,the infrared small targets detection experiment is carried out in 4 kinds of background,and the experimental results show that the proposed algorithm has low false alarm rate and good robustness,comparing with other 5 kinds of commonly-used detection algorithms.