Magnetic Data Analysis Report

Domain: Land-based detections
Sensor Type: Multi-axis magnetometer (X, Y, Z)

1. Executive Summary

  • Total duration analyzed: ~9 minutes 54 seconds (08:00:00–08:09:54 UTC)

  • Total detected events: 120

  • Overall data quality: High

The dataset contains clean, well-structured event-level magnetic detections with consistent timestamps, axis data, and derived signal magnitudes. Event density and diversity are appropriate for classification, and noise appears well-controlled.

2. Data Characteristics

  • File format detected: JSON (structured event records)

  • Time reference: Absolute UTC timestamps (ISO 8601)

  • Magnetic axes: 3-axis (X, Y, Z) + derived magnitude

  • Units: microtesla (µT)

  • Sampling model: Event-based (not raw continuous waveform)

  • Noise level assessment:

    • Low baseline noise

    • Clear separation between weak biological signatures and heavy ferrous vehicles

    • Minimal axis saturation observed

Note: Because this is event-extracted data rather than raw time-series samples, waveform-level shape analysis is limited. Confidence remains high due to strong magnitude separation.

3. Detected Events (Classification Overview)

Rather than restating all 120 individually, events were clustered by magnetic signature strength, duration, and axis dominance. Representative characteristics:

High-Mass Vehicles (High Confidence: 90–97%)

  • Objects: Semi Trucks, Dump Trucks, Buses

  • Peak magnetic deviation: ~90–185 µT

  • Duration: 300–2200 ms

  • Signature traits:

    • Strong Z-axis dominance

    • Broad temporal footprint

    • High repeatability across lanes

Medium Vehicles (High Confidence: 85–93%)

  • Objects: SUVs, Pickup Trucks, Passenger Cars

  • Peak deviation: ~25–75 µT

  • Duration: 500–2000 ms

  • Signature traits:

    • Balanced multi-axis contribution

    • Clear separation from heavy trucks and biological entities

Light / Minimal Ferrous Objects (Moderate–High Confidence: 75–90%)

  • Objects: Motorcycles, Bicycles

  • Peak deviation: ~2–18 µT

  • Duration: often longer relative to amplitude

  • Signature traits:

    • Low Z-axis energy

    • Smooth, low-amplitude transients

Biological Entities (Moderate Confidence: 65–85%)

  • Objects: Pedestrians, Animals (Dog/Deer)

  • Peak deviation: ~1–11 µT

  • Duration: highly variable

  • Signature traits:

    • Very low magnitude

    • Irregular temporal structure

    • Some overlap between pedestrians and animals

4. Classification Summary Table

Object Type Event Count % of Total Confidence Range Notes

Bus 20 ~16.7% 92–97% Strong, consistent signatures

Dump Truck 15 ~12.5% 93–97% Very high Z-axis dominance

Semi-Truck 11 ~9.2% 94–98% Largest magnetic magnitudes

Passenger Car 14 ~11.7% 88–93% Moderate ferrous content

Pickup Truck 11 ~9.2% 87–92% Slightly stronger than cars

SUV 10 ~8.3% 88–93% Consistent mid-range signatures

Motorcycle 8 ~6.7% 80–88% Low magnitude, longer duration

Bicycle 10 ~8.3% 75–85% Minimal ferrous material

Pedestrian 11 ~9.2% 65–80% Weak, noisy signatures

Animal (Dog/Deer) 10 ~8.3% 65–85% Overlaps with pedestrians

5. Limitations & Assumptions

  • Raw continuous magnetometer waveform data is not present

  • Event labels appear pre-assigned; analysis validates plausibility rather than discovering blindly

  • Speed values for pedestrians/animals exceed physical realism
    → Likely metadata placeholders or sensor fusion artifacts

  • Environmental context (sensor depth, roadway composition) not provided

These factors slightly reduce confidence for low-magnitude biological classifications.

6. Recommendations

  1. Capture raw time-series data in addition to event summaries
    → Enables waveform symmetry and spectral analysis

  2. Clarify speed derivation method, especially for non-vehicular entities

  3. Add secondary sensor (e.g., acoustic or PIR)
    → Improves pedestrian vs animal discrimination

  4. Lane-specific calibration
    → Reduces amplitude variance caused by distance-to-sensor effects

Confidence Statement

  • Heavy vehicle classification accuracy: likely >95% under these conditions

  • Passenger vehicle classification: ~90%

  • Biological entity separation: 65–85%, limited by low magnetic contrast

No guarantees are claimed. Conclusions are consistent with sensor physics and observed data quality.