【摘要】 针对单视角像机的视野局限性以及连续跟踪特定目标时运动目标实时检测的困难,提出了一种离散化活动像机运动的方法。先离散化像机运动,建立像机预置位表和背景索引表;然后,利用目标位置信息和运动信息判断像机转动情况,并结合对应的控制原理,实现了像机的调度;最后,利用通过计算机视觉方法确定了特定目标在不同离散空间的对应关系,实现了对特定目标大范围的主动跟踪和运动目标的实时检测。实验结果表明,对于复杂场景中特定运动目标大范围的主动跟踪具有较好的鲁棒性和实时性,并且可以实时提取场景中的运动区域。更多还原

【Abstract】 This paper presents a novel way for object manipulation using an active camera with pan and tilt control to overcome disadvantages of single camera view and the difficulties of real-time detect moving objects in active camera.There are three stages.The first stage is preprocessing to discretize the actively moving camera according to the surveillant scene and create an index table of camera pose,camera parameters and background,which resolve the difficulty of moving objects real-time detection because of moving camera.Then,the desired viewing angle from objects of interest is allotted by using object position and moving information,and the active camera is rotated by corresponding control theory.Finally,location of moving object in different camera pose is based on the computer vision with real-time and accurate capabilities,which adaptively tracks objects of interest in different camera pose.Results on real image sequences illustrate the real-time and robust of the proposed method in large scene.更多还原