Deviation In The Multi-sensor Information Fusion Registration Study

The bias registration in Multi-sensor data fusion is an important branch of multi-sensor information fusion technology. It solves the problem of the declining of the tracking performance through time registration and space registration, which caused by the location error, measurement error, and different sampling time of each sensor. It is a prerequisite of multi-sensor information fusion. So, a strong real-time, high precision and wide application registration algorithm is preferred. According to this, this thesis focuses on the study on time registration and space registration.First, aim at the different sampling rates, gives a real-time registration algorithm, which based on the theory of Lagrange interpolation algorithm and the adaptiveα-βfiltering algorithm. And simulation results show the effectiveness of the algorithm.Second, aim at the space registration, derives ASPF and UCMKF bias estimation algorithms under three-dimensional model. And discusses the ASEKF, ASUKF, ASPF, UCMKF bias estimation algorithms, the simulation results show their different characteristics and apply ranges. Point out ASUKF algorithm has a comprehensive performance in anti-linear, real-time, high filtering accuracy.Then, for large sampling points of ASUKF bringing bigger bias estimation and lower accuracy, an improved ASUKF bias registration algorithm has been advanced. This method effectively reduces the dimension of UT transform and improves the accuracy of real-time. Further study with feedback bias registration algorithm, this method improves each sensor's bias estimation by feedback multi-sensor's estimates and the estimated error covariance. Simulation results show the effectiveness of the algorithm.Finally, design the ASUKF bias registration algorithm under incomplete measurement. Simulation results show the effectiveness of the algorithm, the registration error and target state estimation error decrease as the detection probability rise. The filter is stability when the detection probability bigger than the certain critical value.