为了进一步提高复杂背景下低信噪比红外运动目标的检测性能, 提出了一种基于区域生长和背景配准的低信噪比红外运动目标检测算法。该算法首先利用局部对比度找出几个种子点; 然后根据种子点进行区域生长, 统计生长出目标的相关特征; 接着采用多帧关联的方法, 挑选出候选目标; 最后采用背景配准的方法, 剔除虚假目标, 降低虚警概率。采用大量实验数据进行仿真验证, 仿真结果表明, 相对于传统算法, 本文算法对低信噪比红外运动目标具有较高的检测概率, 较低的虚警概率。

Being aimed to improve the performance of moving infrared target detection against low SNR background, a novel method for moving infrared target detection based on region growing and background registration strategies was introduced. In the proposed method, a few seeds were first searched based on local contrast, which can be used for region growing, then the association characterizes of the growing target was calculated. Next, the candidate targets were selected based on multi-frame association strategy. Lastly, background registration method is used to remove the false targets from the candidate targets set. The experiments show that the proposed method outperforms the traditional methods for moving infrared target detection in low SNR case.