Study On Multi-scale Data Fusion Algorithm And Its Applications

The fusion of the information on multiple scales from multiple sensors can not only achieve better performance than a single sensor,but also describe a better essential characteristic s of the target compared with the fusion on a single scale.MEMS gyroscope is a kind of sensor that can measure angular velocity and has many attractive advantages.But it is also true that it is noisy and inaccurate.Therefore,how to remove the noise in MEMS gyroscope and improve its accuracy has become a research hotspot.Using multi-scale data fusion algorithm for multiple MEMS gyros can significantly improve the accuracy and reliability of the system.In this thesis,the effectiveness of a multi-scale data fusion algorithm which is proposed by predecessors,is proved.A new multiscale fusion algorithm is proposed and the key techniques in multiscale data fusion are discussed.Then this algorithm is used to process the MEMS gyro signal,and the advantages of the multi-scale fusion algorithm are verified by simulation and hardware experiments.The main innovations and works are as follows:1.Based on the wavelet analysis theory,the data fusion theorem under stationary and non-stationary conditions is proved.The principle of multi-scale data fusion algorithm is better than that of classical weighted algorithm is explained mathematically,which lays a mathematical foundation for the popularization and application of this algorithm.2.Combining with the principle and specific implementation steps and existing problems of multi-scale data fusion algorithm,The multi-scale data fusion algorithm using wavelet packet is designed,the comparison between the two kinds of fusion algorithms are simulated with the measured gyroscopes data.3.The selection methods of wavelet basis,decomposition layers and weighting factor in multi-MEMS gyro data fusion are analyzed,and it's feasibility is verified by simulation experiment.4.The performance of the three fusion methods based on time series analysis,wavelet denoising and the multi-scale fusion using wavelet transform is compared with the simulation and measured data.In addition,the multi-scale fusion and the Forward Linear Prediction(FLP)fusion methods are also compared,the results show that the multi-scale fusion method proposed in this thesis is unique and effective.The above research results are applied to a multi-MEMS gyroscope data fusion real-time processing system platform,which is designed and manufactured by us.The original data collected by 4 MEMS gyroscopes are processed in real time.The integrated system is tested in static and dynamic environments respectively.The experimental results show that the system is stable and reliable,and the precision of MEMS gyro is improved by one order of magnitude.The research work in this thesis,which not only lays a theoretical foundation for the analysis of multi-scale fusion system,but also provides an experimental evidence for the popularization and application of the algorithm.