BP neural network based continuous objects distribution detection in WSNs

WSNs (Wireless Sensor Networks) are widely applied in environment monitoring. Especially, in large scale environment monitoring, its flexibility in deployment and self-organization are strong points. However for distribution detection of continuous objects in large scale environment monitoring, there are two primary constraints: energy consumption and the accuracy of the detection which relies on the density of the WSNs. Currently, almost all of the continuous object monitoring are based on the boundary detection, and all the energy efficiency solutions only focus on the WSNs itself. Unfortunately, with the boundary detection method, the accuracy of the continuous objects detection highly relies on the density of the sensor nodes. What is worse, it is even impossible to make sure of the density of the sensor nodes in real situation. In order to deal with these issues, we proposed the Optimal Fusion Set based Clustering algorithm based on the continuous characteristics of the targets to enhance the energy efficiency and Global Distribution Status Monitoring (GDSM) algorithm to implement the monitoring with finite sensor nodes. Firstly, a dynamic diffusion model based on the Gaussian Puff model is proposed, and then the characteristics of continuous objects are analyzed. According to the theoretic analysis and simulation results, the GDSM algorithm can achieve stable accuracy with limited sensor nodes.