Skip to content

Edge Tracking

Edge tracking matches the geometric contours of a 3D model against edges found in the camera image. It uses both the object's outer boundary and internal surface edges (creases, sharp features) for robust tracking.

Edge Tracking Visualization

When to Use

Works well for:

  • Large machinery, engines, and equipment with lots of internal geometric detail
  • Objects with sharp edges, creases, and well-defined geometry
  • Partial views of large objects — internal edges remain visible even when the outline isn't fully in frame
  • Scenes with cluttered or textured backgrounds — edges are more discriminative than silhouettes
  • Low-contrast environments — works with weaker visual contrast than silhouette tracking
  • CAD models — the edge structure matches reality closely

Struggles with:

  • Smooth, featureless objects (spheres, smooth organic shapes)
  • Objects where real edges don't match the 3D model (deformable objects, loose cables)
  • Very small objects with few detectable edges
  • Fast camera or object movement — more sensitive to motion than silhouette

Configuration

On the TrackedBody component:

  • Tracking Method — Select Edge to enable

Edge tracking works directly from the mesh geometry — no tracking model generation required. Just assign the mesh and start tracking. The default parameters work well for most cases.

Edge vs. Silhouette

Aspect Silhouette Edge
What it tracks Object outline vs. background Surface edges + outline
Setup Requires pre-generated tracking model Works directly from mesh — no model generation
Background sensitivity High — needs contrast Low — works in cluttered scenes
Object requirements Visible outline Geometric features (edges, creases)
Motion tolerance Very resilient More sensitive to fast motion
Best for General objects, fast motion Large machinery, partial views, low contrast

Quality Metrics

Edge tracking quality is based on:

  • Coverage — What fraction of the expected edges are successfully matched. The most important factor — drops when edges leave the frame or become occluded.
  • Alignment — How closely the matched edges align with the model. Lower error = better tracking.

Tips

  • Ensure your 3D model accurately represents the real object's geometry — edge tracking relies on geometric precision
  • For objects with both smooth areas and sharp edges, consider combining edge + depth modalities