Research On Parameter Estimation Methods For Sensors' Precision

With the rapid development of information technology,multi-sensor information fusion technology has received increasing attention and applied to complex systems in various fields.It is significant to obtain the accurate value of sensor precision parameters in multi-source fusion system for the parameters' selection within data association and combination algorithm,and so on.However,it is always a challenge to get the accurate value of sensor precision parameters by only using sensors' measuring datum on targets without true values or high-level measuring values.According to the inherent redundancy of multi-sensors' measurement on common targets,an on-line estimation method for sensor precision parameters is put forward.Based on statistical properties of statistical vector,the parameters related objective functions are constructed,and the improved genetic algorithm is used to optimize the objective function.Simulation experiments are conducted to analyze various factors affecting the evaluating effect.The main works are as follows:(1)By analyzing the measurement model of the multi-platform and multi-sensor fusion system,the concept of statistical vector is proposed to describe the statistical relationship of multi-sensor measuring data.It is proved that the statistical vector fits to the standard normal distribution when the parameters estimated are consistent with the preset conditions.Based on the statistical properties and corresponding parameter estimation methods,the objective function is constructed,which can convert the sensor's accuracy estimation problem into a function optimization problem.(2)According to the analysis of the objective function value's distribution graph,the genetic algorithm is selected to solve the optimization problem based on sensor precision estimation.The standard genetic algorithm are improved.The probability of crossover and mutation is converted from fixed values to adapted values and elitisim selection strategy is added.The simulation experiments under matlab environment are performed.The results demonstrate that the improved genetic algorithm performs a better convergence and more accurate estimation than standard genetic algorithm in the distance precision and dirction precision of sensors.(3)We analyze the factors influencing the precision of sensor estimation,including the method of objective function designed,the data volume of measurement data,and the distribution of targets.By perfoming multiple sets of simulation experiments and contrasting the resluts,we get a more favorable scenario for the sensor's precision estimation.The research results show that the estimation method based on statistical vector,statistical test and optimization can carry out an on-line and effective estimation of sensors' precision.The more measuring datum,the more accurate the precision of the estimation results..