Star Detection Engines

indi-allsky tracks celestial clarity and transparency by counting detected stars across captured nighttime frames. The star count is logged in the database (IndiAllSkyDbImageTable), rendered as image overlay text, and plotted on telemetry graphs.

The detection method is configured under Configuration → Processing in the Star & Meteor Detection Engine section.

---

1. Detection Algorithms

A. SEP (Source Extractor in Python) — Recommended

SEP calculates a dynamic, two-dimensional model of the sky background and RMS noise across the frame, subtracts the background, and identifies stellar sources exceeding a statistical significance threshold.

• Advantages: Normalized against shifting background illumination (such as twilight, urban light pollution, and moon phases). Star counts remain mathematically comparable over time.
• Lens Distortion Resilient: Accurately detects elongated or distorted stars near the peripheral edges of wide fisheye lenses.

B. Template Matching (OpenCV)

Template matching cross-correlates image features against a fixed circular star kernel.

• Limitations: Highly sensitive to sky brightness shifts. Because fisheye lenses produce significant radial optical distortion (coma, astigmatism) near the horizon, stars outside the central sensor area are frequently missed.
• Retained for legacy installations or low-power embedded processors where background modeling overhead must be minimized.

---

2. Configuration Reference (SEP Engine)

Parameter	Type	Default	Recommended Range	Description
DETECT_STARS_METHOD	Select	sep	sep / template	Active star detection algorithm.
DETECT_STARS_SEP_THOLD	Float	5.0	3.0 – 15.0	Significance threshold (in standard deviations $\sigma$ above the local background RMS noise). Lower values detect fainter stars but increase susceptibility to noise artifacts.
DETECT_STARS_SEP_MAX_RADIUS	Integer	20	5 – 30	Maximum allowable object radius (in pixels). Excludes large saturated star blooms, satellites, or illuminated cloud edges from inflating star counts.

Calibrating Sigma Thresholds (DETECT_STARS_SEP_THOLD)

• 3.0: Faint limiting magnitude; captures marginal stars but may count high-gain noise spikes.
• 5.0 (Default): Balanced for typical suburban/rural skies.
• 10.0 – 15.0: Restricts detection strictly to prominent navigational stars (useful in light-polluted urban environments).

---

3. Interactive Detection Tuning

Before committing detection parameters:

1. Establish an accurate Detection Mask to exclude trees, buildings, and ground illumination.
2. In the web interface, open Configuration → Image Processing → Detection Tuning.
3. Toggle Run Detection and select Process.
4. The preview image renders detected stars (with centroid bounding boxes) and reports total counts. Adjust thresholds interactively until counts accurately reflect genuine stars.