结合UKF和小波变换的改进粒子滤波及其在机车驾驶员人眼跟踪中的应用

An Improved Particle Filter Algorithm Combining UKF and Wavelet Transform and Its Application to Locomotive Driver Eye Tracking

非接触式的人眼跟踪研究对于机车驾驶员疲劳检测有着重要意义。为了解决机车驾驶员人眼跟踪方法对头部旋转、光照变化以及人眼运动的强非线性等问题的过于敏感,提出一种结合UKF(Unscented Kalman Fil-ter)滤波和小波变换的改进粒子算法的机车驾驶员人眼跟踪方法。将UKF滤波算法得到的滤波状态均值和方差,用于粒子滤波算法中下次采样新的粒子;然后利用小波变换的去噪原理,降低粒子滤波重要性权值的方差,以提高实际驾驶条件下机车司机人眼跟踪的准确性和鲁棒性。理论分析和实验结果表明:该方法不仅可以提高对头部旋转、光照变化以及人眼运动的强非线性等问题的鲁棒性,而且具有更好的估计精度。

Nonintrusive methods for eye tracking are important to locomotive driver fatigue detection. But one common problem to eye tracking methods proposed so far is their sensitivity to head rotation, light variations and eye nonlinear estimation and tracking. In this paper, we present a novel locomotive driver nonlinear eye tracking method based on the improved particle filter algorithm combining UKF and wavelet transform. In order to improve the accuracy and stability of eye tracking, we use the UKF to generate the proposal distribution for the PF （Particle Filter）. Then, we reduce the variance of important weights of the above particle filter using wavelet transform because of the wavelet transform having a good property of denoising, which can improve the accuracy and robustness of eye tracking under realistic driving conditions. The experimental results and theoretical analysis show that the proposed method achieves higher estimation accuracy and robustness of eye tracking to head rotation, light variations and non-linear estimation in realistic driving conditions.