Processing Data

The hybrid_detector.py script analyzes your downloaded InSAR data to identify significant ground motion changes. This guide explains all detection algorithms and processing options.

Basic Usage

Bash
python hybrid_detector.py --path /path/to/downloaded_data --moderate

Detection Algorithms

Choose one algorithm based on your analysis requirements:

Ultra-Selective (`--ultra-selective`)

Default algorithm - Identifies only the most dramatic changes.

Ultra-selective detection for high confidence results
python hybrid_detector.py \
    --path /data/insar \
    --ultra-selective \
    --cpu-threads 8

Best for: - Critical infrastructure monitoring - High-confidence change detection - Reducing false positives

Moderate (`--moderate`)

Balanced approach providing reliable results with reasonable sensitivity.

Moderate detection - recommended starting point
python hybrid_detector.py \
    --path /data/insar \
    --moderate \
    --cpu-threads 4

Best for: - General-purpose analysis - First-time users - Balanced sensitivity vs. reliability

Gradual (`--gradual`)

Detects subtle and gradual changes over time.

Bash
python hybrid_detector.py --path /data/insar --gradual

Best for: - Long-term subsidence monitoring - Slow environmental changes - Research applications

Maximum (`--maximum`)

Extremely sensitive - detects all possible changes.

Bash
python hybrid_detector.py --path /data/insar --maximum

Best for: - Comprehensive surveys - Research requiring all potential signals - Areas with expected subtle changes

Baseline (`--baseline`)

Relaxed criteria for baseline detection studies.

Bash
python hybrid_detector.py --path /data/insar --baseline

Original (`--original`)

Legacy algorithm for backward compatibility.

Bash
python hybrid_detector.py --path /data/insar --original

Processing Parameters

Performance Options

`--cpu-threads`

Number of CPU threads for parallel processing.

Bash
1	`--cpu-threads 8`

Thread Selection

Use 50-75% of your available CPU cores for optimal performance without system overload.

`--coherence`

Minimum coherence threshold for data quality filtering.

Bash
1	`--coherence 0.7`

Coherence Guidelines: - 0.8: Very strict (high quality only) - 0.7: Standard (recommended) - 0.6: Relaxed (includes more data) - 0.5: Permissive (may include noisy data)

Temporal Options

`--fixed-2022-cutoff`

Use fixed date cutoff (2022-01-01) instead of flexible period detection.

Bash
python hybrid_detector.py --path /data/insar --moderate --fixed-2022-cutoff

When to use: - Specific event analysis (pre/post 2022) - Comparative studies - Known temporal boundaries

Algorithm Comparison

Algorithm	Sensitivity	False Positives	Processing Time	Use Case
Ultra-selective	Very Low	Minimal	Fast	Critical monitoring
Moderate	Medium	Low	Medium	General analysis
Gradual	High	Medium	Medium	Subtle changes
Maximum	Very High	High	Slow	Comprehensive surveys
Baseline	Medium-High	Medium	Medium	Research baseline

Output Files

Processing generates two main files:

`change_detection_results.csv`

Contains all detected changes with detailed metrics:

Column	Description
`point_id`	Unique identifier for each point
`longitude`	Point longitude
`latitude`	Point latitude
`change_magnitude`	Magnitude of detected change (mm)
`change_date`	Estimated change occurrence
`coherence_avg`	Average coherence value
`confidence`	Detection confidence score
`velocity_before`	Velocity before change (mm/year)
`velocity_after`	Velocity after change (mm/year)

`file_mapping.json`

Maps internal point IDs to original data filenames for traceability.

Examples

Standard Urban Analysis

Bash
python hybrid_detector.py \
    --path /data/oslo-analysis \
    --moderate \
    --cpu-threads 4 \
    --coherence 0.7

High-Sensitivity Environmental Study

Bash
python hybrid_detector.py \
    --path /data/environmental-study \
    --gradual \
    --cpu-threads 8 \
    --coherence 0.6

Infrastructure Monitoring

Bash
python hybrid_detector.py \
    --path /data/infrastructure \
    --ultra-selective \
    --coherence 0.8 \
    --fixed-2022-cutoff

Interpreting Results

Change Magnitude

< 5mm: Small changes, possibly noise
5-15mm: Moderate changes, investigate further
15-30mm: Significant changes, likely real
> 30mm: Large changes, high confidence

Confidence Scores

> 0.8: High confidence detection
0.6-0.8: Medium confidence
0.4-0.6: Low confidence, verify manually
< 0.4: Very uncertain, possibly false positive

Performance Optimization

Large Datasets

Use more CPU threads (--cpu-threads)
Start with ultra-selective algorithm
Process in smaller spatial chunks
Increase coherence threshold for speed

Memory Considerations

Large areas may require significant RAM
Monitor system resources during processing
Consider processing subregions separately

Troubleshooting

Common Issues

No changes detected - Try a more sensitive algorithm (--gradual or --maximum) - Lower coherence threshold (--coherence 0.6) - Check data quality and coverage

Too many false positives - Use stricter algorithm (--ultra-selective) - Increase coherence threshold (--coherence 0.8) - Check data quality in problematic areas

Slow processing - Reduce CPU threads if system becomes unresponsive - Process smaller areas - Use ultra-selective algorithm for faster results

Memory errors - Process smaller spatial regions - Increase system swap space - Use fewer CPU threads

Next Steps

After processing, proceed to Visualization to explore your detected changes using interactive maps and time-series plots.