基于局部二元模式算子的红外弱小目标检测

Infrared Dim Targets Detection Based on Local Binary Pattern

文中针对在传统红外弱小目标检测中,需要进行背景抑制滤波所带来的图像性质改变和检测速度不理想的问题,提出了一种基于局部二元模式(local binary pattern,LBP)算子的红外弱小目标检测方法。该方法对传统LBP算子进行了改进,使其提取的LBP编码值可以有效地描述红外弱小目标的灰度分布特性,达到了在不进行背景抑制滤波的条件下有效检测弱小目标的目的。结合改进的LBP算子和红外弱小目标灰度的"尖峰"特征,建立了灰度自适应快速扫描机制,有效提高了检测速度,降低了重复告警的出现概率。通过实录红外图像序列检测实验,证明本文方法在检测性能和检测速度方面的有效性和优越性。

In this paper, we present a novel method for infrared dim targets detection based on local binary pattern （LBP）. The modified LBP operator is proposed, by which the extracted LBP values can effectively characterize the intensity distribution of the infrared dim targets. Hence, the reasonable representation model of infrared dim target is built and the good performance of detecting dim targets without background suppressing can be achieved. Combined modified LBP with the intensity feature characteristic of dim target, the fast gray-level adaptive scan strategy is developed to speedup the detection process and restrain the false positive. The experiments result on actual infrared image sequences show that the presented method is feasible and effective.