卡尔曼滤波单目相机运动目标定位研究

Research on Kalman filtering in moving target location with monocular camera

单目相机运动目标定位是视觉跟踪的基本任务之一,视觉传感器对目标点位置的获取至关重要。根据单目相机针孔模型,将运动目标图像二维坐标,通过几何关系映射为相机坐标系的三维坐标。标定相机高度和旋转参数,在此基础上求得目标位置坐标。当运动目标被遮挡时,以匀变速运动为模型,根据先前位置信息确定加速度和速度,并对加速度进行自适应更新,建立运动目标状态方程,利用卡尔曼滤波算法预测目标位置。以预估位置代替真实位置,继续估计后续运动状态,在预估位置搜索运动目标,实现运动目标遮挡定位跟踪。实验结果表明,该定位方法具有可行性,当运动目标发生遮挡时,能够完成对目标的定位和跟踪。

Monocular camera moving target positioning is one of the basic task of visual tracking,visual sensor is the essential part for the acquisition of the target point position. According to the monocular camera pinhole model,the two-dimensional coordinates of the moving object image are mapped to the three-dimensional coordinates of the camera coordinate system according to geometric relation. The camera height and rotation parameters are calibrated,on this basis,the target position coordinates can be calculated. When the moving target is blocked,taking the uniformly variable speed motion as a model,the acceleration and velocity can be determined according to the previous position information and the acceleration is adaptively updated. The state equation of the moving target is established and Kalman filtering algorithm is used to predict the position of the target. The estimated position is used to replace the real position,the subsequent motion state is estimated and the moving target is searched at the estimated position so that the positioning and tracking of the blocked moving target can be realized. Experiment results show that this method is feasible and the moving target can be located and tracked when it is blocked.