在动态场景图像序列中检测运动目标时，如何消除因摄影机运动带来的图像帧间全局运动的影响，以便分割图像中的静止背景和运动物体，是一个必须解决的难题。针对复杂背景下动态场景图像序列的特性，给出了一种新的基于场景图像参考点3D位置恢复的图像背景判别方法和运动目标检测方法。首先，介绍了图像序列的层次化运动模型以及基于它的运动分割方法；然后，利用估计出的投影矩阵计算序列图像中各运动层的参考点3D位置，根据同一景物在不同帧中参考点3D位置恢复值的变化特性，来判别静止背景对应的运动层和运动目标对应的运动层，从而分割出图像中的静止背景和运动目标；最后，给出了动态场景图像序列中运动目标检测的详细算法。实验结果表明，新算法较好地解决了在具有多组帧间全局运动参数的动态场景序列图像中检测运动目标的问题，较大地提高了运动目标跟踪算法的有效性和鲁棒性。

Moving object detection is a very significant and difficult problem in processing image sequence acquired by a mobile camera. The objective of this research is to present a novel method to resolve the moving target detection difficulties, which bring by several layers with different motion parameters of the sequence images. Considering a moving camera, it was hard to distinguish different motion layers led by 3D depth difference of immobile objects from those caused by target moving in scene. Firstly, a mixture model of image was proposed, and the method of motion segmentation based this model were suggested. Secondly, the reconstruction method of scene depth, which employed the motion parameters by EM, was presented. Finally, the detail theory and algorithm of new moving target detection approach, which were based on the scene 3D depth reconstruction, were described. The results indicate that the moving targets detected by the new technique are more accurate than those detected by traditional methods. In addition, the detection speed becomes more-much. The results seem to suggest that the new method be able to provide efficiency and accuracy advantages in computer version.