Next-Best-View Artifact Scanning 2D

Scan Controller

Scanning…

Candidate viewpoints 10

Sensor field of view 30°

Coverage threshold 0.15

Scan Progress

Coverage

Points captured

0 / 0

Viewpoints visited

0 / 0

Sensor angle

0°

Sensor / active viewpoint Unvisited candidate viewpoint Captured boundary point

How it works

This is a flattened, top-down analogue of the arm-and-turntable scanning rig: instead of an elevation-band sweep driven by a spinning turntable, a sensor hops between a fixed ring of candidate viewpoints around a 2D artifact contour. Every boundary point carries an outward normal; a point is captured only if it both falls inside the sensor's field-of-view cone and faces back toward the sensor.

Visibility test for boundary point p, viewpoint v:
  aim   = normalize(-v)              (toward contour centre)
  view  = normalize(p - v)
  angle = acos(view · aim)
  toSrc = normalize(v - p)
  visible = angle < FOV/2  AND  (normal_p · toSrc) > threshold

Next-best-view choice (greedy set cover):
  v* = argmax over unvisited viewpoints v of
       | { points visible from v that are still unscanned } |

After each hop the controller marks that viewpoint visited and jumps straight to whichever remaining viewpoint would reveal the most still-unscanned boundary — the identical greedy "next-best-view" heuristic the 3D rig uses to pick its next elevation band. Too few candidate viewpoints, too narrow a field of view, or too strict a coverage threshold can all leave gaps that no remaining viewpoint can fill, exactly as a real scan-planning system can stall short of full coverage.

⚠ Couldn't load the engineSomething went wrong initializing the canvas. Reload the page.