Passive localization based on radio tomography images with CNN model utilizing WIFI RSSI

Area of Science:

• Computer Science
• Electrical Engineering
• Signal Processing

Background:

• Passive localization is essential for Internet of Things (IoT) applications, especially in private settings like healthcare and security, where individuals avoid wearing tracking devices.
• Indoor environments pose challenges for Global Positioning System (GPS) due to signal degradation (wall loss), necessitating alternative localization methods.
• Radio Tomography Images (RTI) offer a promising approach for passive localization, but traditional methods face limitations due to Radio Signal Strength Indication (RSSI) data imperfections.

Purpose of the Study:

• To develop and evaluate a passive localization system leveraging deep learning and Radio Tomography Images (RTI) for accurate indoor object tracking.
• To address the inverse problem in RTI by utilizing Convolutional Neural Networks (CNNs) for improved image reconstruction and object localization.
• To enhance the quality of tomographic images and achieve high localization accuracy without requiring users to wear tracking equipment.

Main Methods:

• Implementation of a mesh network using ESP32 nodes to form a radio frequency sensor network for collecting RSSI values.
• Development and examination of radio tomography generation algorithms.
• Application of two distinct Convolutional Neural Network (CNN) models to reconstruct static tomographic images and localize targeted objects.

Main Results:

• The proposed system successfully reconstructs static tomographic images and improves their quality.
• The targeted object localization accuracy using the developed CNN models consistently exceeds 92%.
• Comparative analysis demonstrates that the proposed passive localization system significantly outperforms previously developed approaches.

Conclusions:

• The integration of deep learning, specifically CNNs, with Radio Tomography Images (RTI) provides a robust solution for passive indoor localization.
• The developed system offers a high-accuracy, non-intrusive method for tracking objects in environments where traditional GPS is ineffective.
• This research advances passive localization techniques for Internet of Things (IoT) applications, paving the way for more sophisticated monitoring and security solutions.