邱雪娜, 刘士荣. 基于序贯检测机制的双目视觉运动目标跟踪与定位方法[J]. 机器人, 2011, 33(2): 181-190.
引用本文: 邱雪娜, 刘士荣. 基于序贯检测机制的双目视觉运动目标跟踪与定位方法[J]. 机器人, 2011, 33(2): 181-190.
QIU Xuena, LIU Shirong. Moving Objects Tracking and Localization with Binocular Vision Based on Sequential Detection Mechanism[J]. ROBOT, 2011, 33(2): 181-190.
Citation: QIU Xuena, LIU Shirong. Moving Objects Tracking and Localization with Binocular Vision Based on Sequential Detection Mechanism[J]. ROBOT, 2011, 33(2): 181-190.

基于序贯检测机制的双目视觉运动目标跟踪与定位方法

Moving Objects Tracking and Localization with Binocular Vision Based on Sequential Detection Mechanism

  • 摘要: 为了提高复杂环境下的目标跟踪精度,提出了一种基于序贯检测机制的双目视觉运动目标跟踪方法.该方法在序贯检测机制下,将粒子滤波、稀疏场主动轮廓和CamShift等方法结合.首先用基于颜色特征的粒子滤波估计最优跟踪窗口;通过跟踪窗口和目标的相似度决定是否采用稀疏场主动轮廓方法,然后由目标轮廓和目标的相似度决定是否需要CamShift对轮廓进行修正;最后结合双目视觉的视差信息和标定模型对目标进行定位,引入视差置信区间判据可有效减少噪声影响,提高运动目标定位精度.实验表明本文所提的基于序贯检测机制的目标跟踪方法在摄像机运动一目标运动模式下,在目标有尺度、旋转、视角变化和环境有光照变化等情况下,能对运动目标进行有效地跟踪与定位,并且具有比较好的跟踪和定位精度.

     

    Abstract: To improve the accuracy of object tracking in a complex environment,a binocular vision moving object tracking method is presented based on sequential detection scheme.The proposed method integrates particle filter,sparse field active contour and CamShift algorithms under sequential detection mechanism.Firstly,color-feature-based particle filter is adopted to estimate optimal tracking window.Then the similarity between object and the tracking windows determines whether sparse field active contour algorithm should be performed,whereafter the similarity between the contour and the object determines whether CamShift should be employed to modify the object contour.At last,object's location is obtained based on binocular disparity information and calibration model,and the disparity confidence interval criteria is introduced to decrease the impact of the noise effectively and enhance the object localization accuracy.Experiments demonstrate that the proposed method can effectively track and locate the moving object in both camera moving and object moving,and it is able to track and locate the object accurately in both target changes in scale,orientation,view and environment illumination changes.

     

/

返回文章
返回