基于改进梯度抑制的非平面场景运动目标检测

Moving object detection in non-planar scenes based on improved gradient suppression

针对非平面场景3D结构产生的视差对导弹及无人机等空中机动成像平台运动目标检测影响,从提高运动目标检测的快速性和准确性角度,基于梯度抑制算法,提出了改进算法。首先,在非平面场景的运动目标检测中,考虑到视差干扰主要出现在背景边缘上,采用梯度抑制法滤除视差像素。其次,对经过梯度抑制法滤波后的像素,应用外极几何约束,进一步滤除机动成像平台大角度机动产生的剩余视差像素,从而提取出离散的运动目标像素,并进行二值化处理。然后,采用行列投影的方法对二值化的运动目标像素进行分类,确定运动目标区域。最后,对所提的方法进行了实验验证,验证了算法的有效性。

The parallax caused by the 3D structure of the non-planar scenes has a bad influence on the moving object detection from moving imaging platforms such as missiles and unmanned flying vehicles.In order to detect the moving objects in non-planar scenes rapidly and effectively,a novel algorithm combining gradient suppression with epipolar constraint is proposed.Firstly,the gradient suppression algorithm（GSA）is adopted to filter the parallax occurring on the edge of the background.Secondly,as the GSA cannot remove the parallax completely in the case of large angle camera rotation,the epipolar constraint is applied to the remaining pixels to separate the parallax from them.Finally,in order to obtain the boundary of the moving objects,the detected pixels of the moving objects are classified by the projection method of binary images.The experiment results show that the proposed algorithm has a good effect on the moving object detection in non-planar scenes,where the moving objects are all detected and their boundary are identified completely.