PYTHON · NUMPY

Find the Target

Detect a bright target in a noisy image and report its pixel centroid, the input a tracker or a controller acts on.

01Challenge

A downward camera sees a noisy grayscale frame with one bright target. Find where it is — the sub-pixel centroid — robustly enough that background noise doesn't pull the estimate off the target.

02Model

Perception starts by reducing an image to a decision. The simplest useful one: where is the bright thing? Threshold the image to separate target from background, then take the intensity-weighted centroid of what's left — the center of mass of the bright pixels. Noise adds stray bright specks; a threshold set above the noise floor keeps them out so the centroid lands on the target, not the clutter. This centroid is exactly what a tracking loop or a visual-servo controller consumes.

03Guided practice

Worked: build a synthetic frame with a Gaussian target plus noise. Faded: threshold and compute the weighted centroid. Independent: pick a threshold that rejects the noise.

04Feedback

PASS WHEN The thresholded weighted centroid locates the target to sub-pixel accuracy despite noise.

If your run fails, check:

FAIL: centroid pulled off-target — threshold too low, background noise is being counted.
FAIL: no bright pixels pass — threshold too high; lower it below the target peak.

05Retrieve & space

Why weight by intensity instead of a plain pixel average? (Brighter target pixels should count more, giving a sub-pixel center.)
What breaks a single global threshold? (Uneven lighting; then you need a local/adaptive threshold.)
What does this centroid feed? (A tracker's next-frame search, or a visual-servo controller's error term.)

06Mastery & project

Detects a target and reports a sub-pixel centroid robust to background noise (L2: reduce an image to an actionable state estimate).

The centroid is the measurement the Module-3 estimator fuses and the multi-agent search reports as a detection.

← Close the Loop Estimate Motion →

PYTHON · NUMPY

Find the Target

Threshold a noisy frame and compute the target centroid.

You read

the arrays and values already in scope

You change

the code you write in each cell

Fixed

the dataset and the checks that grade you