基于深度信息与码本建模的运动目标检测

Moving object detection based on depth information and codebook modeling

运动目标检测易受阴影干扰,检测时易将阴影误检为运动目标。针对上述问题,文中提出一种基于深度信息与码本建模的运动目标检测方法,首先使用码本建模法得到场景内待确认的目标前景,再通过双目视觉获得场景的深度信息,并根据前景与背景的深度信息差排除阴影的干扰,实现运动目标的检测。实验结果表明,该方法可以较好克服阴影带来的检测难点,并能准确提取出运动目标。

Moving object detection is susceptible to shadow interference,and the shadow is easily detected as a moving target. Aiming at the above problems,this paper proposes a moving object detection method based on depth information and codebook modeling. Firstly,it used the codebook modeling method to get the target foreground to be confirmed in the scene. Secondly,it got the depth information of the scene through binocular vision. Finally,the interference of the shadow is eliminated according to the depth difference between the foreground and the background. The experimental results show that the method can effectively overcome the interference of motion shadow,and can accurately extract the moving target.