视频测量中人工标志点的目标跟踪

Targets Tracking of Artificial Targets in Videogrammetry

针对视频测量中不同时刻标志点的对应问题,提出了一种基于轨迹预测的目标跟踪方法。将二维图像标志点的目标跟踪问题转换为三维空间的多目标跟踪问题,建立了运动模型,从状态方程求取观测方程,用卡尔曼滤波器预测标志点在下一时刻的轨迹,根据实测位置与预测轨迹的接近程度确定对应。论文方法避开了二维图像跟踪所存在的诸多困难,提高了标志点目标跟踪的效率和鲁棒性。

Aiming at the correspondence problem of artificial marks at different times, a matching method based on trajectory prediction was proposed. The 2D mark points is transformed into 3D mark points, mark points＇ kinematic model and motion smoothness cost function are constructed, their trajectories in next frame are predicted using the Kalman filter. According to consistency between actual location and pre-dicted location, correspondences in adjacent frames are established.