NETWORK OF INTEGRATED SENSORS FOR IDENTIFICATION AND TRAJECTORY PREDICTION OF MOVING OBJECTS

Abstract

The paper presents the concept of a distributed network of integrated sensors designed for the identification and trajectory prediction of moving objects using IoT. A method for constructing a spatio-temporal activity field based on acoustic and thermal sensor data integrated within GNSS-equipped nodes is proposed. The study develops a procedure for detecting local anomalies (“hot spots”) in noise and temperature fields using the weighted kriging method, which enables accurate data reconstruction under irregular sensor placement. Spatial–temporal clustering and analysis of signal maxima are performed using the DBSCAN/ST-DBSCAN algorithms, while short- and long-term trajectory forecasting is carried out with Kalman filters (EKF, UKF) and recurrent neural networks (LSTM, GRU). The architecture of an energy-efficient sensor node is described, implementing adaptive data acquisition, preliminary signal processing, and selective data transmission when threshold values are exceeded. The advantages of using LoRa Mesh technology for ensuring self-healing, interference resilience, and autonomous data transmission in distributed monitoring systems are discussed. The proposed approach significantly enhances the reliability and accuracy of object detection over large areas, while reducing the deployment and operational costs compared to traditional radar and optical monitoring systems