Robust Sensing-Hole Detection Integrating with Rotation-Invariant Coordinate Convolutional Neural Networks in Sparse Wireless Sensor Network

Nowadays, the Internet of Things (IoT) is growing rapidly. IoT needs to handle efficient information to handle intelligent applications. There are many applications for Wireless Sensor Networks (WSNs), which act as an important interface between the physical world and the IoT. The continuous object boundary detection is one of the most widely used WSN applications in industry. Traditional WSNs are utilized to track leakage from continuous objects require costly long-term maintenance. To address this issue, sparse WSNs are employed. However, the inaccuracy of sparse networks poses a significant problem when using information about continuous objects to arrange retreat paths for individuals. In this paper, a Robust Sensing-Hole Detection Integrating with Rotation-Invariant Coordinate Convolutional Neural Networks in Sparse Wireless Sensor Network (RSHD-RICCNN-SWSN) is proposed. Initially, fill sensing holes by identifying their virtual points and calculating their variation rates. Then Rotation-Invariant Coordinate Convolutional Neural Networks(RICCNN) is used for detecting the boundary of continuous objects. Finally, Giza Pyramid Construction Optimization Algorithm (GPCOA) is employed for optimal path selection. The proposed RSHD-RICCNN-SWSN method is implemented in MATLAB. The performance measures including Space Significance, Data Significance, Network Lifetime, Energy Consumption, Success Rate, Time Complexity, Detection Accuracy, Network Throughput, and Scalability is evaluated. The proposed RSHD-RICCNN-SWSN method attains 20.58%, 26.73% and 24.62% higher Network Lifetime, 20.98%, 26.66% and 16.32% higher Data Significance, 10.78%, 20.47%, and 12.86% lower Energy Consumption when compared with existing models: Localization and detection of multiple attacks in wireless sensor networks using artificial neural network (LDMA-WSN-ANN), Genetic algorithm for cross-layer-based energy hole minimization in wireless sensor networks (GACL-EHM-WSN), Q-learning dependent vegetation evolution for numerical optimization and wireless sensor network coverage optimization (QLVE-WSN-NOC)respectively.