基于地理同名点配准的机载红外移动小目标检测方法

New method for airborne infrared moving and dim targets detection based on the geographical corresponding points registration

在飞行器等移动载体条件下的红外光电系统中,由于场景条件复杂,目标特征不明显,弱小移动目标检测是自动目标检测领域的一大难点。因此,针对复杂地面场景条件下的移动小目标检测问题,提出了一种无关图像内容的基于帧间地理同名点区域配准的机动目标检测方法。该方法通过综合飞行器惯性组合导航信息,解算视场中像素点对应的地理位置,获取帧间图像地理同名点对,完成图像配准工作,区域配准获得帧间背景的移动量,并对帧间图像进行运动补偿,然后进行帧间差分将背景干扰去除,突出运动目标,最后利用目标运动信息的反向验证对目标进行确认。通过试验结果可以看出,该算法能够取得较好的检测效果。

Target characteristics is not obvious for infrared opto-electronic systems due to the complex scene under the conditions of mobile carders such as aircrafts, as a result in this situation weak moving target detection is a big difficulty in the field of automatic target detection. A new method was proposed for this problem which was independent of image content. Through the combination of aircraft＇s inertial integrated navigation information, the geographical location of the pixel points in the image was calculated and then the corresponding points between the consecutive frames were obtained. So the movement of the background was acquired and the motion compensation was conducted, and the target was highlighted. Finally, the target was confirmed by backward verification using target motion information. The experiments show that the new algorithm can obtain a better detection results.