Table of Contents
⚠ Disclaimer: This entry may be incomplete, out of date, or inaccurate. It is AI-maintained on a best-effort basis. Do not rely on it as a sole source — verify claims independently using the sources listed below.
Summary
eye_sky is an open-source Python system for camera-based multi-object tracking across multiple camera feeds, with a binocular camera package supporting 3D triangulation. It originated as a Python port and extension of SPOTIT3D, a MATLAB system developed by Dr. Sutthiphong Srigrarom, adding filtering, visualization, and additional triangulation capability. Two peer-reviewed papers grew out of this line of work: one on continuous drone tracking through camera-to-camera “frame-stitching” as an object crosses between fields of view, and one on binocular camera systems for multi-drone detection specifically.
Key Facts
- Author: Yi-Jiahe (GitHub); builds on SPOTIT3D by Dr. Sutthiphong Srigrarom (originally MATLAB)
- Type: Open-source software — multi-camera, multi-object tracking with 3D triangulation
- Status: Small/active — 3 stars, 0 forks as of this review
- Core stack: OpenCV, NumPy, FilterPy, SciPy, Matplotlib
- Associated publications: Yi, Chew & Srigrarom (2021) on continuous drone tracking via frame-stitching across camera transitions; Yi & Srigrarom (2020) on binocular camera systems for multi-drone detection
How It Works
The system combines background subtraction and real-time detection with an object-tracking pipeline that handles video I/O and image transformations for multiple simultaneous camera feeds. An optical-flow-based camera stabilization module compensates for camera motion/jitter before tracking. The “Binocular Camera” package pairs two camera views to perform 3D triangulation, directly analogous in goal (if not implementation) to this section’s Stereo Drone Tracker. The project structure includes separate modules for real-time multi-camera tracking, binocular 3D triangulation, multi-view indoor tracking, comparisons of different object trackers, and integration scripts for DJI Tello drones (likely used as a convenient, controllable test target).
Relevance to This Section
Dr. Sutthiphong Srigrarom’s broader body of work on trajectory-based drone-vs-bird classification (motion-characteristic-based discrimination rather than appearance-based) is directly in the same research lineage documented in this section’s TRAMIS entry, which cites a related Srigrarom paper (“Drone versus bird flights: Classification by trajectories characterization”) among its references. eye_sky and TRAMIS independently arrive at a similar architectural thesis — multi-camera, trajectory/track-based processing rather than per-frame image classification — for the same underlying problem.
Limitations
- Small project, thin documentation on current state: No confirmed license, last-commit date, or contributor count found in available material — verify directly on GitHub before relying on or building against it.
- No consolidated performance metrics in the main repository; benchmark/accuracy figures would need to be sourced from the associated Yi/Chew/Srigrarom papers rather than the code repo itself.
- Research-grade code: Like most projects in this section’s open-source list, this is more likely to need adaptation (calibration, synchronization, outdoor robustness) than to be production-deployable as-is.