One key challenge for sensor networks is to provide the real-time high reliable sensor measurements with the minimum resource consumption. Outlier clearance in sensor networks can ensure the quality of sensor measurements and dependable monitoring. In this paper, we propose two online distributed outlier clearance approaches with low computational complexity and memory usage that can identify and remove the spurious sensor measurements. The proposed approaches are operated locally and thus save communication overhead as well as possess good scalability. The evaluation performance of proposed approaches and existing widely used methods on synthetic and real-life dataset illustrates that our Adaptive Top-n WAD approach achieves remarkable outlier clearance performance as compared to these methods.

传感器网络的一项关键挑战是以最少的资源消耗提供实时，高可靠性的传感器测量。传感器网络中的异常间隙可以确保传感器测量的质量和可靠的监视。在本文中，我们提出了两种在线分布式离群值清除方法，这些方法具有较低的计算复杂度和内存使用率，可以识别和消除寄生传感器的测量值。所提出的方法在本地操作，因此节省了通信开销并具有良好的可扩展性。在合成和现实数据集上，所提出的方法和现有广泛使用的方法的评估性能表明，与这些方法相比，我们的自适应Top-n WAD方法具有显着的离群清除性能。