【作者】 李雪松；

【导师】 勇俊；

【作者基本信息】 哈尔滨工程大学 ， 水声工程， 2019， 硕士

【摘要】 声学多普勒测速声纳是利用多普勒效应原理,精确测量平台对地绝对或对水相对速度的一种重要海洋观测仪器,已广泛应用在海洋资源开发、导航控制、航道测量等诸多领域;随着作业需求的不断提高,精确测速问题一直受到人们的关注。本文着重围绕未知非平稳测速噪声中的速度估计问题,从自适应滤波处理角度研究了提高测速估计精度的方法。声呐性能受环境影响严重,而未知非平稳噪声假设更符合声学多普勒测速声呐的噪声模型。基于“厚尾”噪声分布的速度估计或鲁棒性速度估计等方法解决了经典卡尔曼滤波的不足,然而,由于参数需要人为设计,实际应用具有一定的局限性。本文基于变分贝叶斯方法研究噪声统计模型,实现对未知非平稳测速噪声的实时推断,在此基础上,利用变分贝叶斯卡尔曼自适应滤波算法(VBAK,Variational Bayes Adaptive Kalman Filter)实现对速度信息的最优估计。我们从以下两个角度评估VBAK算法性能:1)构建了两个典型测速场景,并通过直接性指标(RMSE、CEE%、EE%)和间接性指标(估计误差与真实误差相关系数)对VBAK算法进行了定量评估;2)通过处理外场试验数据评估了VBAK算法的实际应用效果。结果表明:VBAK算法相比于经典Kalman滤波算法优势明显,能够较好地解决未知非平稳噪声情况下的声学多普勒测速信息估计问题,估计精度得到进一步提高,达到了预期目标。更多还原

【Abstract】 Acoustic Doppler velocity sonar is an important ocean observation device that accurately measures the bottom/water three-dimensional velocity using the Doppler Effect.It has been widely used in marine resources development,ship navigation,flow measurement and many other fields.The continuous improvement of the demand for work,the problem of accurate speed measurement has been concerned by people.In this paper,we focus on the speed estimation problem in unknown non-stationary velocity measurement noise,and study the method of improving the accuracy of velocity measurement from the perspective of adaptive filtering.The sonar performance is severely affected by the environment,and the unknown non-stationary noise assumption is more in line with the acoustic model of the acoustic Doppler velocity sonar.The method of velocity estimation or robustness velocity estimation based on the "thick tail" noise distribution solves the shortcomings of the classical Kalman filter.However,since the parameters need artificial design,the practical application has certain limitations.In this paper,based on the variational Bayesian method,the noise statistical model is studied to realize the real-time inference of unknown non-stationary velocity measurement noise.Based on this,the variational Bayes Adaptive Kalman Filter(VBAK)is used.Achieve an optimal estimate of velocity information.We evaluate the performance of VBAK algorithm from the following two aspects: 1)Construct two typical velocity measurement scenarios,and quantitatively evaluate VBAK algorithm through directness index(RMSE 、 CEE% 、 EE%)and indirect index(estimation error and true error correlation coefficient);2)The actual application effect of the VBAK algorithm is evaluated by processing the field test data.The results show that the VBAK algorithm has obvious advantages over the classical Kalman filtering algorithm,and can solve the problem of acoustic Doppler velocimetry information estimation under unknown non-stationary noise conditions.The estimation accuracy is further improved and the expected target is achieved.更多还原