基于中值滤波和小波变换的混合滤波方法在推进剂监测中的应用

目前,监测传感器传出信号中混有很多噪声,为提高信号可信度,需要一种有效的信号处理方法。文章基于Matlab仿真环境,完成了信号仿真和滤波算法的设计,重点对单传感器仿真信号的去噪和多传感器信息融合进行了研究,提出了基于中值滤波和小波阈值滤波的混合滤波方案和基于Kalman滤波的信号融合方案。研究工作有:基于高斯白噪声和脉冲噪声的数学特性,合理假设出5种基本信号形式;依据实际数据,完成单传感器和多传感器信号仿真,确定信噪比和均方根误差作为去噪评定指标;综合分析现有的滤波算法的滤波特性,利用不同长度滑动窗口的中值滤波处理实验信号,选取合适长度的滑动窗口。设置对比实验确定小波阈值滤波中的小波基函数选取、阈值计算和分解尺度等参数;融合中值滤波和小波阈值滤波优势,设计混合滤波方案,去除单传感器仿真信号中的噪声;研究信息融合理论在泄漏监测系统中的应用,设置不同融合方式下的对比实验,确立最佳融合方式下的Kalman滤波方案,实现多传感器信息融合。

The Application of Hybrid Filtering Method Based on Median Filtering and Wavelet Transform in Propellant Monitoring

At present, a reliable signal processing method is needed because the signal of monitoring sensor is mixed withstrong noise. Based on simulation environment of Matlab, the signal simulation and filtering algorithm are designed in thispaper. The emphasis is signal denoising and the information fusion of multi-sensor. A hybrid filtering scheme based on me?dian filtering and wavelet threshold filtering and a signal fusion scheme based on the Kalman filtering are proposed. Belowis the research work: Five basic signal forms are reasonably assumed based on mathematical characteristics of Gaussianwhite noise and impulse noise. According to actual data, the signal simulation of single sensor and multi-sensor is complet?ed. In the meantime, signal-noise ratio and root-mean-square error are used as the evaluation indicators of denoising.Having analyzed existing filtering algorithms, the appropriate length of the sliding window is selected when using middlefiltering to process signal. At the same time, a contrast experiment is set up to solve the problem of selecting the wavelet ba?sis function, calculating the threshold and determining the decomposition scale in the wavelet threshold filtering. Combin?ing the advantages of median filtering and wavelet threshold filtering, a hybrid filtering scheme is designed to achieve thedenoising of single sensor. Considering the application of information fusion in leakage monitoring system, experiments ofchoosing best fusion modes and the scheme of Kalman filtering are set up so as to realize the information fusion of multi-sensor.