红外多光谱图像弹道目标检测算法

在红外多光谱图像中,弹道导弹尾焰拥有两大特征,一是由强烈红外辐射引起的灰度差异,二是独特的光谱特性。然而,传统的单波段检测技术只利用了尾焰强烈的辐射特性,而近些年发展起来的多光谱检测技术则只利用了尾焰独特的光谱特性。为了充分利用导弹尾焰的两大特征,将单波段检测技术和多光谱检测技术结合起来,提出三种检测算法,并从算法的检测效果、运算量和鲁棒性三方面详细分析它们的优缺点。采用人工合成的红外多光谱图像进行验证,实验结果表明,相比单独使用单波段或多光谱的检测算法,融合算法的检测性能更好。

Ballistic target detection in infrared multispectral imagery

There are two important characteristics for missile plume in infrared multispectral imagery, one is the gray-scale difference caused by strong infrared radiance, and the other one is the unique spectral signature feature. However, the classical single-band detection technology only uses the first characteristic, and the multispectral detection technology which has developed in recent years only uses the second characteristic. In order to fully exploit the characteristics of missile plume, we propose three detection algorithms by combining the single-band and multispectral detection technology. The advantages and disadvantages of the three algorithms are discussed in detail on detection performance, computational complexity, and robustness. Finally, experiments on synthetic infrared multispectral imagery demonstrate the better performance of combined algorithms when compared with single-band or multispectral detection algorithm.