基于粒子滤波器的摄像机运动参数估计

针对利用运动矢量估计摄像机运动参数的问题,将运动矢量作为状态,以块的位置作为状态索引,根据区域的局部相关性自适应地确定状态转移方程,以块匹配的平均绝对误差作为状态估计的观测输出,从而建立状态空间模型。基于该模型利用粒子滤波器估计运动矢量,采用重采样技术来消除粒子退化的现象,最后按照运动矢量与摄像机参数的关系模型,借助最小二乘法确定最优的摄像机参数。实验表明,该算法以较少的搜索点可以估计得到具有时空一致性的运动矢量场,进而能估计出反映摄像机真实运动的摄像机运动参数。

摄像机运动模式分类方法研究

在动态场景下进行运动目标检测,面临的首要问题就是如何来消除摄像机运动造成的影响。因此研究摄像机运动模式是很有必要的,论文对摄像机的运动模式进行了归纳分类,总结了三种主要类型的运动矢量场分布特点,最后给出了分类方法和分类流程。

Simultaneous Position Estimation and Omnidirectional Camera Parameter Calibration for Multiple Mobile Robots

This paper proposed an algorithm on simultaneous position estimation and calibration of omnidirectional camera parameters for a group of multiple mobile robots. It is aimed at developing of exploration and information gathering robotic system in unknown environment. Here, each mobile robot is not possible to know its own position. It can only estimate its own position by using the measurement value including white noise acquired by two omnidirectional cameras mounted on it. Each mobile robot is able to obtain the distance to those robots observed from the images of two omnidirectional cameras while making calibration during moving but not in advance. Simulation of three robots moving straightly shows the effectiveness of the proposed algorithm.