基于Kalman滤波的头盔运动姿态的预测方法

Pose prediction method for moving helmet based on Kalman filtering

针对头盔瞄准具单目视觉定位系统中,由于面阵CCD相机帧频的限制而导致对其定位速度慢的问题,提出了一种基于自适应Unscented Kalman滤波的头盔运动姿态固定超前预测方法。首先,分析了飞行员头部的运动特点,利用匀变速运动来描述头盔的实际运动,并建立了头盔姿态固定超前预测的状态空间模型;然后,基于自适应Unscented Kalman滤波理论分别对头盔在相机连续2次采样中间的非采样时刻和采样时刻的姿态进行固定超前预测,并利用采样时刻头盔姿态的测量值对其固定超前预测值进行实时的修正,以在间接提高头盔定位速度的同时,保证其定位精度;最后,搭建了单目视觉姿态测量与预测实验系统,对该方法的有效性和准确性进行了验证。实验结果表明,该方法能够对头盔瞄准具在相机连续2次采样中间的非采样时刻的运动姿态进行有效而准确的固定超前预测,在保证了对头盔定位精度的前提下,可以有效地将其定位速度提高1倍。

Aiming at the problem of slow speed of Helmet-Mounted Sight＇s positioning, which is restricted by the area CCD camera＇s frame rate in monocular vision positioning system, a fixed-time prediction method for moving helmet pose based on adaptive Unscented Kalman filtering is proposed in this paper. Firstly, the moving characteristics of the pilot＇s head is analysed, and the helmet＇s actual movement is described by the uniformly variable motion, then the state-space model is established to describe the fixed-time prediction process for moving helmet pose; Secondly, the helmet poses are fixed-time predicted based on adaptive Unscented Kalman filtering, on both the camera＇s sampling time and the non-sam- piing time that is between two continuous sampling times, and then the predicted value on the sampling time is modified by the measuring one, so that the helmet＇s positioning speed can he improved indirectly, at the same time, its positioning accuracy would not decline; Finally, to verify the validity and veracity of the proposed method, the experimental monocu- lar vision pose measurement and prediction system is built. The experimental results show that the pose of Helmet- Mounted Sight on CCD camera＇s non-sampling time can be predicted exactly by the proposed method, so the positioning speed of helmet can be effectively doubled, on the premise of ensuring its positioning accuracy.