GPS Map Matching — Score-Field View (2D)

Trace Controls

GPS noise σ 3.0 m

Vehicle speed 6.0 u/s

Heading weight 0.45

Distance tolerance σ_d 2.2 m

Live Readouts

Raw GPS scatter (avg)

0.0 m

Matched-point error (avg)

0.0 m

Correct-road rate

GPS fixes taken

How it works

This 2D view renders the map-matcher's decision field directly: every point of the plane is coloured by which road segment currently wins the scoring function, and how brightly, instead of only showing the single winning trace. It's the same algorithm as the 3D version, visualised as a live classifier landscape rather than a driving camera.

foot = clamp_project(p, a, b)
d    = |p - foot|
distScore    = exp(-d² / 2σ_d²)
headingScore = |dot(dir(a,b), heading)|
continuity   = 1.0 same road as last fix, 0.6 shares a junction, 0.15 otherwise
score = distScore · (1 − w_h + w_h·headingScore) · (0.4 + 0.6·continuity)

Each road segment gets its own hue; the field's brightness at a pixel is that segment's score there. Drag the map to pan, scroll/pinch to zoom. Raw noisy fixes (red) and the matched trace (blue) are overlaid on top of the field so you can see exactly where the classifier boundary caught — or missed — the true road.

GPS noise σ — scatter of a raw fix from the true position.
Heading weight — how much compass course counts vs. raw proximity.
Distance tolerance σ_d — how sharply the field's brightness falls off away from a road's centreline; small σ_d gives thin bright bands, large σ_d floods whole blocks with colour.

Raw noisy GPS fix Map-matched trace Vehicle (true position)