基于多尺度特征融合的红外图像小目标检测

针对复杂背景红外图像弱目标检测问题,提出了一种基于多尺度特征融合的红外小目标检测方法。该方法在对红外图像进行离散小波框架变换的基础上,提取目标在各子带图像上的多个特征;然后对各子带图像上的多个特征分别进行融合,获得相应的目标检测判决置信度图;接着采用自适应加权方法对各个目标检测判决置信度图进行融合,得到目标检测判决总置信度图,并对其进行目标区域分割,获得最终的目标检测结果。实验结果显示,该方法对高杂波条件下红外弱目标检测具有良好效果。

IR image small target detection based on multi-scale feature fusion

Aim at the IR image weak target detection problem in challenging condition, A detection method based on multi-scale feature fusion for IR small target is presented. This method firstly extracts multiple features of target in theses sub-band images after transforming the IR images using discrete wavelet frame transform. Then fuses multiple feature images of all different sub-band images independently to get the corresponding target detection decision confidence images. In following, fuses theses sub-band confidence images using adaptive weighting algorithm to get the total target detection decision confidence image and segment the total confidence image to get the final target detection result. The experimental results show that this method's effectiveness for IR small target detection under highly clutter background.