Raster Change Detection

Two co-registered satellite passes over the same ground, weeks apart — what changed? Change Vector Analysis, an automatic Otsu threshold, morphological speckle removal, and connected-component extraction, backed by a native C++ kernel.

C++ Change Kernel

Detection Controls

Scene size

Sensor noise σ 0.025

Threshold mode

Otsu (automatic)

Manual

Manual threshold 0.20

Open iterations

Minimum blob area (px)

Detections — click a row to highlight it on the mask

Multitemporal Stack

–

Detections

–

Changed Area

–

Threshold

–

Ground Truth Recovered

Before / after swipe — drag the divider or use the slider below. False-colour composite (NIR, Red, Green): healthy vegetation reads red.

Divider position

Change mask over epoch B — solid boxes are detections, dashed boxes are planted ground truth.

CVA magnitude histogram — the vertical line is the chosen threshold. Raise the noise and watch it move.