云背景红外弱小目标结构相似检测算法

Matched Structure Approach to Detect Small IR Targets under Cloud Clutter

由于自然物具有结构自相似性的分形几何特征,云背景条件下,天空和云彩的红外图像在灰度分布上具有强的自相似性,而目标与自然物边缘是奇异点。小目标与自然物只有边缘,没有分形结构的自相似性。采用八种分形结构对区域内像素进行旋转映射,确定最小误差为该区域的分形结构,进行图像恢复,恢复出的图像只包含背景。当原始图像与恢复图像的灰度差大于阈值时,检测出的目标为弱小目标。

Seen from the fractal geometry,the material existed in the nature world has similarity and self-similarity in their structures,which leads to the similarity in gray scale distribution of the infrared image,especially under the environment of cloud clutter.According to the features,the fractal estimation method is put forward to separate the targets and background in the cloud clutter IR images.The method,which assumes the target location is polluted,includes a fixed searching location and 8 transforms to set searching templates.After matching calculator,the background can be acquired,and through subtraction between the original image and the background image,the targets can be detected.