Research On Multi-sensor Temporal Registration Problem Based On Gaussian Process Regression

Multi-sensor time registration is a key technology of the front-end data processing of the information fusion system.The quality of time registration is one of the important factorsaffectingperformance of information fusion system.With the rapid development of science and technology in the information field,information fusion systems are more and more widely used in civil and military fields,as the premise of data fusion,time registration technology is indispensable in asynchronous information fusion system.Interpolation extrapolation ? Least square method are commonly used methods of time registration technology,but these methods have the limitations of using conditions and causelarge error of time registration under more complex motion model.To solve these problems,high-performance time registration method has become a research hotspot in the field of information fusion.This paper introduces and analyzes the time registration technology in the information fusion system from three aspects: the basic concept of time registration,the common methods of time registration and the moving model of maneuvering target.According to the principle of time registration and Gaussian process regression method,the application of Gaussian process regression algorithm in the time registration technology of information fusion system is mainly studied,and the concept of remodeling is proposed.The multi-dimensional Gauss random variable is constructed by asynchronous observation value and its time,and the time registration is realized by solving conditional probability distribution.In this paper,the detailed theoretical analysis of time registration using Gaussian process regression is given,and the effectiveness of the algorithm is verified by simulation.This paper proposes to obtain the actual position information of the moving target as sample data under the premise of knowing the target motion model and parameters(process noise);According to the time registration principle and the characteristics of the Gaussian process regression algorithm,select the appropriate kernel function and combine the sample data to solve the kernel function using maximum likelihood estimation Hyperparameters;change the motion model parameters,repeat the steps of solving the kernel function,establish the numerical mapping relationship between the motion model parameters and the Gaussian process kernel function hyperparameters;for asynchronous observation information that requires time registration,use Gaussian process regression Predict and achieve fusion;finally,the computer simulation results verify the effectiveness of the Gaussian process regression time registration method and show that it has higher accuracy than the traditional method.