diagnose-alignment

Diagnose Alignment

Check alignment health and recommend fixes for the active Metashape chunk.

Diagnostic Steps

1. Check project state: get_alignment_stats for overall alignment summary
2. Check cameras: What percentage are aligned? Are there clusters of unaligned cameras?
3. Check tie points: How many exist? What's the ratio of points to cameras?
4. Check sensors: Are calibration parameters reasonable? Use list_sensors
5. Check spatial distribution: get_camera_spatial_stats for coverage gaps

Common Issues and Solutions

Low alignment rate (<80%)

• Try match_photos with higher keypoint_limit (60000–100000)
• Enable guided_matching=True
• Lower downscale (0 or 1 for highest accuracy)
• Check for images with insufficient overlap

High reprojection error (>1 pixel)

• Run optimize_cameras with all distortion coefficients
• Try adaptive_fitting=True
• Remove cameras with very high individual errors

Sparse tie points (<1000 per camera)

• Increase keypoint_limit and tiepoint_limit
• Check image quality with analyze_images
• Remove blurry images (quality < 0.5)

Alignment drift / banding

• Add ground control points (GCPs)
• Use reference_preselection=True
• Check camera GPS accuracy
• See /corridor-alignment-pipeline for incremental approach

Output

Provide a summary of findings with specific recommended tool calls and parameters.