Abstract: The target do irregularity motion in wireless sensor networks, make it difficult to select reasonable data to predict the target future position, meanwhile it is easy to happen target lost during tracking process and go against the target monitoring. By analyzing the change characteristic of target motion trend, the paper puts forward multi-cluster heads structure and target recovery strategy in order to improve the tracking accuracy and reduce traffic lost, then improves linear fitting method so that selects the appropriate data in a certain stage as prediction data. The simulation results show that the improved algorithm can use historical data efficient, and has good performance in tracking and prevent from drooping.

Key words: Wireless Sensor Networks（WSN）, piecewise linear approximation, prediction, target tracking, Least-Square（LS）, localization

摘要： 在无线传感器网络中，目标的运动无规律可循，难以选取合理的数据对其进行预测，且在跟踪过程中易发生目标丢失的现象，不利于对目标的监测。通过分析目标运动趋势的变化特点，提出多级簇头结构和目标丢失的恢复策略，并改进线性拟合算法，选取某一阶段内的数据作为预测的数据源，以提高跟踪精度，降低目标丢失率。仿真结果表明，改进算法能够更充分地利用历史数据，在跟踪精度和目标丢失率方面具有良好的性能。

关键词: 无线传感器网络, 线性拟合, 预测, 目标跟踪, 最小二乘法, 定位