摘要
现有的基于特征点的跟踪方法不能准确估计目标状态,构建多参数动态更新的局部不变特征点数据集,依据自动特征尺度提取理论,利用目标上的特征点尺度变化估计目标大小变化,利用特征点主方向变化估计目标方向旋转;采用相对于目标中心的坐标描述特征点位置,利用目标大小和方向旋转信息校正该坐标并进行目标中心重建;特征点匹配时不仅要求描述符相似,还要求校正后的空间位置一致,可有效去除误匹配。实验证明,所提算法在动、静摄像平台下均适用,能准确定位目标、跟踪其尺度和方向变化,遮挡时也能正确估计目标状态。
Existing tracking algorithms based on feature sets up a multifrom the scale c points cannot estimate object state accurately. This paper parameter dataset of feature points that is updated dynamically. The size of the object is estimated hange of feature points on the object according to the automatic feature scale selection theory, and the rotation of the object is calculated using the orientation change of feature points. The spatial position of feature point is depicted with the coordinates relative to the center of the object, used to correct the relative coordinates and the center of the object the size and rotation information of the object is is reconstructed using the corrected relative coor- dinates. Feature point matching requires that not only the descriptors are similar but also the corrected spatial positions are consistent, so that false matches can be eliminated effectively. Experiments demonstrate that the presented algorithm can be used no matter the camera is in static or moving states, can accurately locate the object and track the scale and orientation changes, and can correctly estimate the object state even under occlusion condition.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2012年第9期2053-2060,共8页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(61105015)
江苏省科技支撑计划(BE2011747)
江苏省自然科学基金(BK2011511)资助项目
关键词
目标跟踪
尺度空间
特征点
模板更新
object tracking
scale space
feature point
template update