Easy Tracker

Simplified point tracking with flexible model configurations

Easy Tracker is a streamlined version of PointTracker with a simplified interface and flexible model configurations. It tracks bright points (typically IR LEDs) arranged in 3, 4, or 5-point models, making it versatile for different hardware setups.

How It Works

Easy Tracker detects bright blobs in the camera image and matches them to a predefined 3D point model. Using OpenCV’s PnP (Perspective-n-Point) solver, it calculates the head’s 6DOF pose. The tracker includes a Kalman filter for smooth, jitter-free output.

Easy Tracker offers multiple PnP solver options and supports custom 3, 4, and 5-point models, making it more flexible than the standard PointTracker.

Requirements

Hardware

• Camera: PS3 Eye camera or webcam with IR filter removed
• IR LEDs: 3-5 infrared LEDs (850nm or 940nm)
• Power source: Battery pack for LEDs (3-5V)
• LED mounting: Cap, clip, or headset for LEDs

Software

• OpenCV (included with OpenTrack)
• Camera drivers

Same hardware requirements as PointTracker, but with more flexibility in LED arrangement.

Setup Instructions

1. Choose Your Model Type

   Easy Tracker supports multiple LED configurations:

   • 3-point model: Standard triangle (cap or clip style)
   • 4-point model: Quad configuration for better stability
   • 5-point model: Maximum accuracy and rotation range
   • Clip model: Specialized 3-point clip configuration

   Select your model type in tracker settings.

2. Build LED Array

   Create your LED array according to chosen model:

   For 3-point model:

   • Arrange 3 LEDs in asymmetric triangle
   • Measure positions relative to center point
   • Record X, Y, Z coordinates in millimeters

   For 4-point model:

   • Arrange 4 LEDs with at least one out of plane
   • Ensure no 3 points are collinear
   • Measure all positions

   For 5-point model:

   • Create extended configuration
   • Use widely spaced points for best results
   • Measure positions carefully
3. Configure Camera

   Set camera parameters:

   Camera: Select your device
   Resolution X: 640
   Resolution Y: 480
   FPS: 30-60
   Field of View: 56 degrees
   
4. Set Blob Detection Parameters

   Adjust blob detection for your LEDs:

   Min Blob Size: 4 pixels
   Max Blob Size: 15 pixels
   

   View the preview to verify only LEDs are detected.

5. Enter Model Coordinates

   Input your measured LED positions:

   For 3-point custom:

   Vertex Top (X, Y, Z): e.g., (0, 50, 0)
   Vertex Right (X, Y, Z): e.g., (40, -30, -20)
   Vertex Left (X, Y, Z): e.g., (-40, -30, -20)
   

   For 4-point:

   Vertex Top (X, Y, Z)
   Vertex Right (X, Y, Z)
   Vertex Left (X, Y, Z)
   Vertex Center (X, Y, Z)
   

   Coordinates are in millimeters, origin is at one reference point.

6. Select PnP Solver

   Choose pose estimation algorithm:

   • SOLVEPNP_P3P: Fast, requires exactly 3 points
   • SOLVEPNP_ITERATIVE: Robust, works with any number of points
   • SOLVEPNP_EPNP: Efficient for 4+ points
   • SOLVEPNP_IPPE: Good for planar configurations

   Default P3P works well for 3-point models.

7. Configure Deadzone

   Set deadzone for steady aiming:

   Deadzone Half Edge Size: 1-5 pixels
   

   Larger values reduce jitter but may feel less responsive.

8. Enable Auto-Center (Optional)

   Configure automatic centering:

   Auto Center: true
   Auto Center Timeout: 1000ms
   

   Automatically centers view when you stop moving for the timeout period.

Configuration Options

Camera Settings

Option	Default	Description
Camera Name	-	Select camera device
Resolution X	640	Width in pixels
Resolution Y	480	Height in pixels
FPS	30	Target framerate
FOV	56°	Camera field of view

Blob Detection

Option	Default	Description
Min Blob Size	4	Minimum blob diameter in pixels
Max Blob Size	15	Maximum blob diameter in pixels

Model Configuration

Option	Default	Description
Custom Model Three	true	Enable 3-point custom model
Custom Model Four	false	Enable 4-point model
Custom Model Five	false	Enable 5-point model
Clip Model Three	false	Enable clip-style 3-point model

Model Vertices (Example for 3-point)

Vertex	X	Y	Z	Description
Top	0	0	0	Top/front LED position
Right	0	0	0	Right LED position
Left	0	0	0	Left LED position

All coordinates in millimeters from origin point.

Solver Settings

Option	Default	Description
PnP Solver	SOLVEPNP_P3P	Pose estimation algorithm
Deadzone Half Edge	1	Deadzone size in pixels
Auto Center	true	Enable automatic centering
Auto Center Timeout	1000ms	Time before auto-center triggers
Debug	false	Show debug information

PnP Solver Comparison

Solver	Points	Speed	Accuracy	Use Case
P3P	Exactly 3	Very Fast	Good	Standard 3-point tracking
ITERATIVE	Any	Medium	Good	General purpose
EPNP	4+	Fast	Very Good	4 or 5 point models
IPPE	4+	Fast	Good	Planar arrangements
SQPNP	Any	Medium	Very Good	When accuracy critical

Advanced Features

Kalman Filter

Easy Tracker includes a built-in Kalman filter that:

• Smooths position and rotation outputs
• Predicts pose during brief point occlusions
• Reduces jitter from blob detection noise
• Automatically adapts to movement speed

The filter is always enabled and tuned for responsive yet smooth tracking.

Multiple Point Models

Support for different point counts provides flexibility:

3-point models:

• Simplest to build
• Lowest computational cost
• Good for most use cases
• May have pose ambiguity at extreme angles

4-point models:

• Eliminates most pose ambiguity
• More robust tracking
• Better for extreme head rotations
• Slightly more complex to build

5-point models:

• Maximum tracking stability
• Best rotation range
• Most robust to point occlusion
• Requires careful LED placement

Vertex Matching

The tracker uses sophisticated vertex matching to identify which detected blob corresponds to which model point:

1. Distance-based matching: Compares blob positions to expected positions
2. Geometry validation: Ensures detected points form valid configuration
3. Temporal consistency: Uses previous frame to resolve ambiguities
4. Fallback strategies: Multiple matching algorithms for robustness

Troubleshooting

Wrong points detected

• Adjust min/max blob size to filter out false detections
• Check for reflections in camera view
• Increase LED brightness relative to background
• Reduce threshold if LEDs too dim
• Block windows/lights causing reflections

Pose jumps or flips

• Verify model coordinates are accurate
• Check that model is asymmetric (not symmetric triangle)
• Try different PnP solver (ITERATIVE more robust)
• Increase blob size constraints
• Ensure LEDs are rigidly mounted

Tracking unstable or jittery

• Increase deadzone size
• Verify LEDs are securely mounted
• Check for camera focus issues
• Use higher camera resolution
• Ensure consistent LED brightness
• Check mounting stability

Solver fails / No pose solution

• Verify at least minimum points detected (3 for P3P)
• Check model coordinates are entered correctly
• Try different PnP solver
• Ensure model points aren’t collinear
• Check that model scale is reasonable

Auto-center not working

• Verify auto-center is enabled
• Check timeout setting (try increasing)
• Ensure you’re actually stopping movement
• Debug mode shows center status

Performance Tips

Optimize Speed

• Use P3P solver for 3-point models (fastest)
• Lower camera resolution (640x480 sufficient)
• Reduce FPS if CPU limited
• Minimize blob size range

Improve Accuracy

• Use 4 or 5-point model
• Increase camera resolution
• Larger LED spacing in model
• Use EPNP or SQPNP solver

Better Stability

• Increase deadzone size
• Use 4+ point model
• Ensure rigid LED mounting
• Use consistent LED brightness

Wider Range

• Use 5-point model
• Space LEDs widely
• Use ITERATIVE solver
• Increase max blob size

Comparison with PointTracker

Feature	Easy Tracker	PointTracker
Point models	3, 4, or 5	3 only
PnP solvers	5+ options	Fixed
Kalman filter	Always on	Optional
Auto-center	Built-in	No
Setup complexity	Slightly easier	Standard
Performance	Very similar	Very similar
Flexibility	More flexible	More mature

Tips for Best Results

1. Model Design: Use asymmetric arrangements - avoid regular triangles or rectangles
2. LED Spacing: Wider spacing = better accuracy, aim for 80-120mm between points
3. Solver Selection: Start with P3P, switch to ITERATIVE if you get pose flips
4. Calibration: Measure LED positions accurately - 1mm error affects results
5. Mounting: Ensure LEDs don’t flex or move relative to each other
6. Testing: Use debug mode to verify point detection before finalizing setup

See Also

• PointTracker - Alternative 3-point tracker
• Hardware Guide - Building LED arrays
• Wiimote Tracker - Alternative IR tracking hardware