Tracking Technology Comparison
OpenPTrack was created to support interactive projects by UCLA REMAP, our collaborators, students, and colleagues.
No one sensing technology is the best fit for every project. REMAP uses a variety of systems for sensing position, gesture, and pose in our workâfrom high end commercial motion capture to inexpensive open source solutions for blob tracking.
By comparing some popular approaches (that we also use), the table below illustrates the gap that OpenPTrack aims to fill: scalable, robust, real-time person tracking using affordable off-the-shelf components and an open source codebase.
OpenPTrack v2 | Single Kinect | Blob Tracking (off-the-shelf) |
Augmented Reality | Motion Capture Marker-based |
Motion Capture Markerless |
RF Tracking | |
(w/ many popular frameworks) | (e.g., Community Core Vision) | (e.g., AR Toolkit) | (e.g., Vicon) | (e.g., Open Perception) | (e.g., Zebra) | ||
Target audience | Education, Arts, Culture | Various | Various | Various | High-end Production | High-end Production | Industrial sensing |
Core technology | Networked 3D imagers | 3D Imager | 2D Camera(s) | 2D Camera | 2D Cameras | 2D Cameras | Radio Frequency |
Output Type | ID, 3D Centroid, Skeletal Data,
Pose Data, Orientation, and Object Data |
ID, Skeletal Data | ID, 2D Centroid | ID, 3D Position, Orientation | ID, Dense Skeletal Data | ID, Skeletal Data | ID, 3D Position |
Max. Tracking Volume | Large | Small | Small to Medium | Small to Medium | Can be very large | Medium | Can be very large |
Fusion of multiple views | Intrinsic | N/A | Up to developer | Up to developer | Intrinsic | Intrinsic | N/A |
Typical refresh rate (Hz) | 30-60 | 30 | 15-30 | 30-60 | 60-120+ | 60-120 | 20-50 |
Lag (perceptual) | Low | Low | Low | Low | Very Low | Low | Medium |
Maximum people tracked | Many | Typically 4 | Tradeoff with volume | Tradeoff with volume | Many | 4 | Many |
Person detection | Yes | Yes | Not usually | N/A | Yes | Yes | N/A |
ID Stability | Medium | Medium to High | Low | High | High | Medium to High | Very High |
3D Tracking | Yes | Yes | No | Yes | Yes | Yes | Yes |
Skeletal Tracking | Yes | Yes | No | No | Yes | Yes | No |
Must carry / wear something | No | No | No | Yes | Yes | No | Yes |
Position accuracy | High | High | Varies greatly | Medium to High | Very high | Very high | Medium |
Occlusion resistance | High in multi-imager nets | Low | Low | Low | High | Medium | Requires multiple tags |
Visible light sensitivity | Minimal | Minimal | Yes | Yes | Some | High | None |
IR light sensitivity | Depends on imager | Yes | Depends on imager | Depends on Imager | Often | Not usually | No |
Costume sensitivity | Must be humanoid | Must be humanoid | None | Must show tag | Must wear marker suit | Must be humanoid | None |
Multiple imager types | Yes | No | Yes | Yes | Yes | Yes | n/a |
Typical setup time | Medium | Very low | Low to Medium | Low to Medium | Medium to High | Medium to High | Medium to High |
App integration complexity | Low | Low | Low to Medium | Low to Medium | Medium to High | Medium to High | Low to Medium |
Open source software | Yes | Usually | Usually | Usually | No | No | No |
Off-the-shelf parts | Yes | Yes | Yes | Yes | No | No | No |
Typical system cost | $$ | $ | $ | $ | $$$$ | $$$$ | $$$ |
(Feel free to email us with additions and corrections to this chart!)