基于DLBP特征的FLIR图像匹配方法

FLIR image matching method based on DLBP feature

针对FLIR复杂地面目标无直接可用基准图、背景干扰严重、目标与背景灰度差异小、不利于匹配识别的问题提出了一种新的方法。在现有高程数据、正射影像、目标区可见光纹理数据的基础上生成目标区场景的参考图像,并在FLIR实时图中检测出候选目标区域。通过分析FLIR图像与可见光图像纹理,可知其具有一定的相似性。因此,基于局部二值模式(LBP)纹理描述子,将参考图像与候选目标区域的红外纹理特征进行匹配,提出了一种基于方向估计的LBP(DLBP)描述方法,使得LBP对于光照和旋转变化更加鲁棒。实验结果表明:该方法对于复杂背景的前视红外目标匹配率高,稳定性好,有效地解决了前视红外图像匹配问题。

According to the FLIR image of complex fixed ground target has no available base image,serious background interference,and small gray differences between target and background,a new method for image matching was introduced.The base image depending on the altitude data,orthophoto map and texture information of target area was created,and then the candidate garget regions from FLIR real image was detected.The texture of FLIR image and optical image have certain similarities.So matching them based on texture feature,and a new LBP descriptor based on Radon direction estimation（DLBP） was proposed,which makes the LBP more robust to illuminate and rotate.The experiment results show that the method has high precision and quick performance,the algorithm has high practicality for FLIR target under complicated background.