Gauge Image Processing System

A computer vision system that automatically reads analog pressure gauges using a Raspberry Pi camera. It captures images of physical gauges, detects the needle position, converts the angle to pressure values (PSI/BAR), and tracks readings over time with graphical analysis.

What It Does

1. Captures gauge images every 20 minutes using a Pi camera with LED lighting
2. Detects the circular gauge face and needle position using OpenCV
3. Calculates the needle angle and converts it to pressure (PSI/BAR)
4. Stores readings in a SQLite database with timestamps
5. Plots time-series graphs showing pressure trends over time
6. Filters outlier readings automatically

Perfect for monitoring pressure systems, boilers, compressors, or any analog gauge that needs logging.

Quick Start

One-Command Deployment to Raspberry Pi

# Clone this repo, then deploy everything:
./deploy_to_pi.sh

# Or specify custom host:
PI_HOST=pi@raspberrypi.local ./deploy_to_pi.sh

This script automatically:

• Installs all dependencies (uv, sqlite3, pigpio, libcamera)
• Copies all required files
• Configures and starts both services
• Verifies the deployment

Manual Processing

# Process captured images and generate plots
uv run gauge_cli.py --plot --pressure-unit bar --average --all-time

# Remove outlier readings
uv run filter_large_angles.py --mark-as-failures

How It Works

System Architecture

Two systemd services work together:

1. dial_capture.service - Image capture

   • Runs continuously, sleeps until next capture time
   • Captures at :00, :20, :40 minutes past each hour
   • Controls LED via GPIO for consistent lighting
   • Saves JPEGs to /home/jack/dial_images/

2. gauge_processor.service - Image processing

   • Checks for new images every 5 minutes
   • Processes images (10-20 min per image on Pi)
   • Deletes processed images to save space
   • Updates database and plots

Image Processing Pipeline

1. Load image and convert to grayscale
2. Detect circular gauge using Hough circles
3. Extract gauge region and apply threshold
4. Detect needle using line detection
5. Calculate angle from vertical (0-360°)
6. Convert angle to pressure using calibration
7. Store results in database

Data Storage

SQLite database (gauge_data.db) with two tables:

• gauge_results - Successful readings (angle, pressure, timestamp)
• detection_failures - Failed detection attempts

Configuration

Edit gauge_config.toml to adjust:

[gauge.detection]
binary_threshold = 140      # Image threshold for needle detection
min_radius = 100           # Minimum gauge size in pixels
max_radius = 1000          # Maximum gauge size in pixels

[gauge.pressure]
min_angle = 31             # Angle at 0 pressure
max_angle = 265            # Angle at max pressure
max_psi = 58              
max_bar = 4.0

Monitoring & Management

# Check service status
ssh jack@pi4light "sudo systemctl status dial_capture.service gauge_processor.service"

# Watch live processing logs
ssh jack@pi4light "sudo journalctl -u gauge_processor.service -f"

# View recent readings
ssh jack@pi4light "sqlite3 gauge_data.db 'SELECT datetime(timestamp), angle, pressure_bar FROM gauge_results ORDER BY timestamp DESC LIMIT 10;'"

# Check disk space
ssh jack@pi4light "df -h /home/jack"

Command-Line Options

gauge_cli.py

Main processing tool with options:

• --plot - Generate time-series plots
• --pressure-unit [angle|psi|bar] - Display units
• --time-window N - Days to include (default: 7)
• --all-time - Plot all historical data
• --average - Enable time-based averaging
• --average-period [minute|hour|day] - Averaging granularity
• --force - Reprocess all images

filter_large_angles.py

Outlier detection:

• --threshold - Maximum valid angle (default: 200°)
• --mark-as-failures - Move outliers to failures table

Alternative: Local Processing

For faster processing on a more powerful machine:

# Windows - Process images locally
.\read_it.ps1

# Windows - Sync database from Pi and view combined data
.\sync_from_pi.ps1

# Windows - Quick view of Pi's current plot
.\view_plot_from_pi.ps1

# Linux/Mac
scp jack@pi4light:./dial_images/*.jpg ./dial_images/
uv run gauge_cli.py --plot --pressure-unit bar --average --all-time

Database Syncing

The sync_from_pi.ps1 script:

• Downloads new readings from the Pi's database
• Merges with your local database (no duplicates)
• Regenerates plots with all historical data
• Opens the updated plot automatically

This allows you to maintain a complete historical record locally while the Pi continues to delete processed images to save space.

File Locations

On Raspberry Pi:

• Images: /home/jack/dial_images/
• Database: /home/jack/gauge_data.db
• Plots: /home/jack/gauge_plots.png
• Scripts: /home/jack/
• Services: /etc/systemd/system/

Troubleshooting

Processing seems stuck:

# Check if actively processing (high CPU usage is normal)
ssh jack@pi4light "ps aux | grep gauge_cli"

# Restart if needed
ssh jack@pi4light "sudo systemctl restart gauge_processor.service"

No images being captured:

# Check capture service
ssh jack@pi4light "sudo systemctl status dial_capture.service"
ssh jack@pi4light "sudo journalctl -u dial_capture.service -n 50"

Database errors:

# Install sqlite3 if missing
ssh jack@pi4light "sudo apt-get install -y sqlite3"

Performance Notes

• Image processing takes 10-20 minutes per image on Raspberry Pi
• CPU usage can reach 300%+ during processing (normal)
• Each image is ~1MB, processed images are deleted automatically
• Database and plots are retained indefinitely

Repository Contents

• deploy_to_pi.sh - Automated deployment script
• capture_images.sh - Camera capture with LED control
• continuous_gauge_processor.sh - Processing loop
• gauge_cli.py - Main image processing tool
• gauge_lib.py - Core detection algorithms
• gauge_config.py - Configuration loader
• gauge_config.toml - User settings
• filter_large_angles.py - Outlier detection
• sync_database_from_pi.py - Database synchronization
• sync_from_pi.ps1 - Windows script to sync and view data
• view_plot_from_pi.ps1 - Quick plot viewing
• read_it.ps1 - Local image processing
• Service files and installation scripts