Temperature Hotspot Detection & Tracking

---

System Overview

This repository implements Bayesian Inference, Gaussian Processes, Kalman Filters, and Particle Filters for detecting and tracking temperature hotspots over time and space. The methods are applied to simulated sensor data to identify, estimate, and track hotspot dynamics.

---

High-Level Architecture

Detection & Estimation Layer

• Bayesian Inference: Probabilistic hotspot detection using prior and likelihood models
• Gaussian Processes (GPR): 2D spatial temperature estimation with smooth interpolation

Time-Series Tracking Layer

• Kalman Filters: Noise reduction and temporal tracking of hotspots
• Particle Filters: Probabilistic evolution of hotspots over time using Monte Carlo simulations

Visualization Layer

• Interactive plots, density plots, heatmaps, and time-series graphs for analysis and monitoring

---

Execution Flow

1. Bayesian Inference

   • Assumes hotspots are rare (prior belief)
   • Updates posterior probabilities with sensor readings
   • Identifies hotspots where posterior mean exceeds a threshold

2. Gaussian Process Regression (GPR)

   • Models temperature field with RBF kernel
   • Predicts temperature over a 2D grid
   • Detects hotspots where predicted mean > 35°C

3. Kalman Filter

   • Filters noisy time-series sensor data
   • Estimates true temperature over time
   • Tracks activation and evolution of hotspots

4. Particle Filter

   • Simulates hotspot evolution probabilistically
   • Resamples particles based on measurement likelihood
   • Predicts future temperature states under uncertainty

---

Results & Visualization

• Density plots: Show probability distributions of temperature readings
• Heatmaps: Highlight hotspot locations and intensities
• Time-series graphs: Compare raw vs. filtered temperature data

Bayesian Inference	Temperature Field Evolution	Detection Performance (ROC)
Real-time Monitoring Dashboard	Method Performance Comparison	Uncertainty Quantification

---

Scalability & Extensibility

• Can integrate real sensor data for live monitoring
• Extendable to multi-dimensional sensor networks
• Supports additional probabilistic or machine learning–based tracking methods

---

Applications

• Environmental monitoring and hotspot detection
• Sensor network data analysis
• Real-time anomaly detection in temperature fields
• Educational demonstrations of Bayesian and probabilistic filtering

---

Contribution

Feel free to fork and contribute to improve models, visualizations, or performance!

---

License

MIT License. Free to use, modify, and distribute for academic and research purposes.