中文摘要：

针对目标发射功率变化下的无线传感器网络（WSN）目标定位问题，分析了无线信道衰减特性，探讨目标功率无关的信号强度差特征提取方法，结合WSN信息交换与处理过程，提出能消除WSN目标功率变化影响的信号强度差LSSVR建模定位方法（TL—LMSD），该方法利用不同探测节点平均信号强度差构造特征向量，通过LSSVR回归建模获得表征特征向量与目标坐标映射关系的LSSVR模型，将各节点目标信号强度测量值的差值所构造特征向量输入LSSVR模型可实现目标定位．基于CC2430无线传感网络实验平台证明TL-LMSD方法目标定位均方根误差RMSE比MLE方法可减小29％～37％；TL—LMSD方法在LSSVR建模、无需重新建模2种情况下的目标定位耗时分别约为0．4s、0．04s．这表明TL-LMSD方法能显著减小信号强度值变化对目标定位结果的影响，提高目标定位准确度，并具有较好的实时性能．

英文摘要：

To better locate targets when the target＇s transmitting power changes, characteristics of wireless channel attenuation were analyzed. Extraction of features of signal strength differences unrelated with the target＇s transmitting power was also examined. A novel target localization method was proposed for wireless sensor networks （WSN） combining information exchange and processing, and based on least squares support vector regression （LSSVR） modeling of signal strength differences. With this method, mean received signal strength differences between different sensors were used to compose the initial feature vector. Then LSSVR models were obtained by LSSVR modeling; they represented mappings of relationships between a feature vector and a target＇s coordinates. Moreover, targets were located by inputting the feature vector composed of the measured signal strength differences among sensors into LSSVR models. The target localization experiment was performed using a WSN experimental platform using Texas Instruments CC2430 chipsets. It was proven that the root mean square error of the target localization-LSSVR Modeling Received Signal Strength Difference （TL-LMRSSD） method＇s maximum likelihood estimation （MLE） error reduced from 29% to 37%. Furthermore, the processing time for the TL-LMSD method was about 0.4s when LSSVR modeling was needed and about 0.04s when LSSVR modeling was not needed. This showed that the TL- LMSD method improved target localization accuracy by significantly reducing the influence of signal strength variations on target localization results. In addition, it showed good real-time performance.