PROJECT TITLE :

Boundary Tracking of Continuous Objects Based on Binary Tree Structured SVM for Industrial Wireless Sensor Networks

ABSTRACT:

In the petrochemical and nuclear industries, the flammability, explosiveness, and toxicity of continuous objects (for example, chemical gas, oil spill, radioactive waste) make boundary tracking of continuous objects an extremely important concern for industrial wireless sensor networks (IWSNs). In this piece, we propose a continuous object boundary tracking algorithm for IWSNs. The algorithm makes full use of the sensor nodes' innate capabilities for collective intelligence and Machine Learning, and it tracks the boundaries of objects in real time. The algorithm that is being proposed begins by establishing a maximum limit for the event region that is covered by the continuous objects. Within the event region, a binary tree-based partitioning procedure is carried out in order to obtain a coarse-grained boundary area mapping. The boundary tracking problem is then transformed into a binary classification problem so that the irregularity of continuous objects can be studied in greater depth. A hierarchical soft margin support vector machine training strategy is designed to address the binary classification problem in a distributed manner. The results of the simulation show that the proposed algorithm demonstrates a reduction in the number of nodes required for boundary tracking that is at least fifty percent lower than the previous algorithm. The algorithm that has been proposed does not require any additional fault-tolerant mechanisms because it is inherently robust to false sensor readings. This holds true even for high ratios of faulty nodes (>9%).