Thermal Data Extraction Tool

Extract temperature data from drone thermal images (DJI and FLIR-based cameras).

Supported Drones

Fully Supported (DJI)

• DJI Mavic 3 Thermal (M3T)
• DJI Mavic 2 Enterprise Advanced (M2EA)
• DJI Matrice 30T (M30T)
• DJI Matrice 300 RTK with Zenmuse H20T/H20N
• All DJI drones with thermal cameras producing R-JPEG files

Experimental Support (FLIR-based)

• Skydio X10 with VT300-Z sensor (FLIR Boson+)
• Autel EVO II Dual 640T
• Yuneec H520E with E10T/E20T
• Parrot Anafi USA/Thermal
• Any drone with FLIR thermal sensors (Boson, Lepton, Tau)

Features

• Extract raw temperature data from thermal images
• Export to CSV format (x, y, temperature)
• Export metadata to JSON
• Generate thermal heatmaps
• Calculate temperature differences between pixels
• No data loss - direct sensor readings

Installation

Prerequisites

For DJI images:

• Docker (required - no alternatives available)
• DJI uses proprietary format that requires official SDK via Docker
• macOS, Windows, and Linux all need Docker

For FLIR/Skydio images:

• Python 3.7+
• exiftool
• Python packages: numpy, matplotlib, pillow
• No Docker needed

Setup

1. Install Docker Desktop (for DJI images)

2. Install exiftool:

brew install exiftool  # macOS
# or
sudo apt-get install exiftool  # Linux

3. Install Python dependencies:

pip3 install numpy matplotlib pillow

4. Build Docker image (for DJI):

docker build -t dji-thermal-extractor .

Quick Start

Extract from DJI Image

# Place your DJI thermal image in the data folder
mkdir -p data/output

# Extract thermal data
docker run --rm \
  -v "$(pwd)/data:/data" \
  dji-thermal-extractor \
  node extract_thermal.js /data/YOUR_DJI_IMAGE.JPG /data/output

Extract from Skydio/FLIR Image

# Place your thermal image in the data folder
mkdir -p data/output

# Extract thermal data
python3 extract_flir_thermal.py data/YOUR_IMAGE.JPG data/output

Output Files

For each image, you get:

1. CSV file (*_thermal_data.csv)

   • Format: x, y, temperature_celsius
   • One row per pixel
   • Ready for Excel, Python, MATLAB

2. JSON file (*_thermal_data.json)

   • Image dimensions
   • Thermal parameters (emissivity, distance, humidity)
   • Temperature statistics (min, max, average)

Analyze Thermal Data

Generate Heatmaps and Statistics

python3 analyze_thermal.py data/output/thermal_data.csv data/output

Output:

• Thermal heatmap (PNG)
• Temperature profile graph (PNG)
• Statistics printed to console

Get Temperature at Specific Pixels

# Single pixel
python3 get_pixel_temp.py data/output/thermal_data.csv 320 256

# Compare two pixels
python3 get_pixel_temp.py data/output/thermal_data.csv 100 100 600 500

Solar Panel Inspection

Automatically detect defective solar panel cells:

python3 solar_panel_inspection.py data/output/thermal_data.csv data/output

Output:

• Segments panels from vegetation
• Identifies problem areas (defective cells)
• Draws red boxes around hotspots
• Exports detailed report with locations

Usage Examples

Solar Panel Inspection

# Extract thermal data
python3 extract_flir_thermal.py data/solar_panel.jpg data/output

# Analyze and create visualizations
python3 analyze_thermal.py data/output/solar_panel_thermal_data.csv data/output

# Find hot spots (anomalies)
# Check the heatmap: data/output/solar_panel_heatmap.png

Building Thermal Survey

# Extract from DJI drone
docker run --rm \
  -v "$(pwd)/data:/data" \
  dji-thermal-extractor \
  node extract_thermal.js /data/building.jpg /data/output

# Get temperature at specific window
python3 get_pixel_temp.py data/output/building_thermal_data.csv 450 320

Batch Processing

# Process all DJI images in a folder
for img in data/*.JPG; do
  docker run --rm \
    -v "$(pwd)/data:/data" \
    dji-thermal-extractor \
    node extract_thermal.js "/data/$(basename "$img")" /data/output
done

# Process all FLIR/Skydio images
for img in data/*.JPG; do
  python3 extract_flir_thermal.py "$img" data/output
done

Verification

The extraction uses official SDKs and proven formulas:

DJI images:

• Official DJI Thermal SDK
• Conversion: Raw value / 10 = Temperature (°C)
• Accuracy: ±2°C or ±2% (per DJI specifications)

FLIR images:

• Standard FLIR Planck formula
• Uses camera calibration constants from image metadata
• Accuracy depends on camera model (typically ±2-3°C)

To verify:

1. Open image in manufacturer software (DJI Thermal Analysis Tool, FLIR Tools)
2. Click on a pixel
3. Compare with extracted CSV data - should match exactly

Troubleshooting

"Failed to create DIRP handle"

Your image is not a DJI format. Try FLIR extraction:

python3 extract_flir_thermal.py data/image.jpg data/output

"Docker is not running"

Start Docker Desktop application before running extraction.

Temperature values seem wrong

Check that:

1. Image is radiometric (R-JPEG), not visual-only
2. Camera thermal mode was enabled during capture
3. You're using the correct extraction method (DJI vs FLIR)

Some pixels show extreme values

For FLIR images, a small number of edge pixels (<1%) may have errors. This is normal and doesn't affect the rest of the data.

Technical Details

DJI Format

• Format: Proprietary R-JPEG with DIRP thermal data
• Raw values stored as uint16
• Temperature (°C) = Raw Value / 10

FLIR Format

• Format: Standard R-JPEG with embedded TIFF thermal data
• Uses Planck's radiation formula for conversion
• Formula: T = B / ln(R1 / (R2 * (raw + O)) + F) - 273.15
• Constants (R1, R2, B, F, O) stored in image EXIF metadata

File Structure

.
├── README.md                    # This file
├── Dockerfile                   # Docker setup for DJI extraction
├── package.json                 # Node.js dependencies
├── extract_thermal.js           # DJI extraction script (runs in Docker)
├── extract_flir_thermal.py      # FLIR/Skydio extraction script
├── analyze_thermal.py           # Analysis and visualization tool
├── get_pixel_temp.py            # Quick pixel temperature lookup
├── data/                        # Input images go here
│   └── output/                  # Extracted data saved here
└── .dockerignore               # Docker ignore file

Python API

Load Thermal Data

import csv
import numpy as np

# Load from CSV
temperatures = []
with open('thermal_data.csv', 'r') as f:
    reader = csv.DictReader(f)
    for row in reader:
        temperatures.append(float(row['temperature_celsius']))

# Reshape to 2D array (height x width)
# For DJI M3T: 1280 x 1024
# For Skydio X10: 640 x 512
temp_array = np.array(temperatures).reshape(height, width)

# Get temperature at pixel (x, y)
temp = temp_array[y, x]  # Note: y first, then x

# Calculate difference between two pixels
diff = temp_array[y2, x2] - temp_array[y1, x1]

Contributing

This tool is provided as-is for thermal data extraction. Feel free to modify and share.

License

Open source - use freely for any purpose.

References

DJI Thermal Extraction

• DJI Thermal SDK - Official DJI thermal data extraction library

FLIR Thermal Calculation Formula

The FLIR temperature conversion uses Planck's radiation law with factory calibration constants.

Official FLIR Documentation:

• FLIR Temperature Measurement Formula - Official formula documentation
• FLIR Radiometric JPEG Format - R-JPEG specification
• FLIR UAS Radiometric Tech Note - Technical documentation for drone cameras

Open Source Implementations:

• Thermimage R Package - Academic research implementation
• flirpy Python Library - Python thermal imaging library

Formula: T = B / ln(R1 / (R2 × (raw + O)) + F) - 273.15

Where R1, R2, B, F, O are Planck calibration constants stored in image EXIF metadata.

Credits

• DJI Thermal SDK by gmazoni
• FLIR Planck formula from official FLIR documentation
• Built with Node.js, Python, Docker

Support

For issues or questions about specific drone models, check the manufacturer documentation:

• DJI: https://www.dji.com/support
• FLIR: https://www.flir.com/support
• Skydio: https://support.skydio.com