系统误差校正是多传感器数据融合和跟踪系统中的基本问题，传统的解决方法是最小均方估计法或极大似然估计法，其缺点是对噪声比较敏感。文中提出了用非参数化方法解决多传感器数据融合中的系统误差校正问题，具体地讲就是把系统误差校正问题转化为非线性优化问题，然后通过模拟退火算法求解。该方法的优点是不需要事先知道各传感器的系统误差，并且适用于不同类型的传感器。经过仿真可知算法有效且对噪声不敏感，比线性化方法有更高的收敛效率和求解精度。

Sensors registration is a basic problem in all mutisensor data fusion and tracking systems. Traditionally, it is solved by the least square method or maximum likelihood method which are sensitive to noise. In this paper, we propose a new nonparametic approach for sensor registration. Firsly, the problem of systematic errors in multiple sensor data fusion is transformed into nonlinear optimization problem. Then, it is solved with simulated annealing algorithm. By numerical simulation, the method of this paper shown to be more efficient and precise than linear method and at the same time it is not sensitive to noise.

TN958