BLOOD TRAIL

Feature weights

D = green band depth, the real discriminator. Q = red/green ratio. S = blue suppression.

D weight.55

Q weight.30

S weight.15

Ranges

Where each feature maps 0 to 1. Widen if nothing lights up, narrow if everything does.

D min0.25

D max0.70

Q min1.90

Q max6.0

S low.05

S high0.35

Spatial

Blood is small and discrete. Local contrast rejects large uniform areas like red clay banks and rust colored leaf beds.

Local contrast.35

Ring radius12

Blood type preset

Different organs bleed differently. Bright is lung, muscle and arterial. Dark is liver, heart and venous. Aged is 6 hours or more, browned out. Wide catches all three and false-alarms more.

WIDE BRIGHT DARK AGED

Gut shot blood mixed with rumen contents is green brown and this will not find it. No hemoglobin index will.

Rejection

Red floor is the single most important control. The index is a ratio, so in dark areas it explodes on sensor noise and anything faintly warm reads as blood. This is why black cables, shadows and dark trim light up. Raise it until they stop.

Red floor.015

SKIN VETO

Skin veto rejects bright pixels with only a moderate band depth and uncrushed blue. Ordinary skin of any tone already scores low. Lips and fresh scrapes still trigger, because they genuinely are hemoglobin.

Response

If the overlay feels slow to light up, the usual cause is the phone auto white balance pulling red back out over about a second. LOCK freezes it. SNAP removes the fade so hits pop instantly.

SNAP EDGE

Analysis res.50

Lower is faster. .50 means the detector runs at half resolution, which is plenty for droplets and roughly four times the frame rate.

Camera

White balance1.00 / 1.00 / 1.00

TAP A GRAY CARD RESET WB LOCK AWB/AE

Point at something neutral gray or white, hit the gray card button, then tap it on screen. This cancels the phone's auto white balance drift, which otherwise moves every number you record.

Display

NIGHT UI DIAGNOSTICS FIELD PROTOCOL

No readings yet. Hit PROBE, tap a spot, pick a label.

Goal of the first session is not to find deer. It is to find out whether the index separates blood from your ground cover. Take readings, send them back.

Setup

1. Get real blood. Butcher shop, or the next deer. A cup is plenty.
2. Drip it, do not pour it. Single droplets, a spatter, and one smear.
3. Put samples on each substrate separately: green grass, dry oak or maple leaves, pine needles, bare dirt, red clay if you have it, gravel, a downed log.
4. Before every set of readings, hit the gray card calibration on something neutral.

For each substrate, PROBE and label

1. One reading on the blood.
2. Three readings on the bare substrate right next to it, no blood.
3. One reading on the worst lookalike nearby: a red leaf, a berry, rust, orange tape.

Conditions to repeat it in

• Open sun and full shade
• Dusk
• Full dark with the torch on. Android torch works, iPhone Safari does not expose it.
• Fresh, then the same spots again at 4 hours and overnight. Aged blood browns out and that is the case that matters.

Then

Open LOG, hit COPY ALL, paste it to me. Grab screenshots too, especially any spot where the overlay lit up on something that was not blood.

What the numbers mean

• D is the one that matters. Modelled: green vegetation about -2.0, gray dirt about 0.00, brown oak leaf +0.13, red clay +0.42, aged blood +0.51, fresh blood +0.79. These are estimates. Your readings replace them.
• Q is redness. Brown leaves about 1.9, clay about 3.0, fresh blood 6 to 9.
• S is blue over red. Blood crushes blue, so lower is more blood like. Weakest of the three.
• Known false positives, do not be alarmed: red maple leaves, berries, rust, hunter orange. Red leaf pigment absorbs green in almost the same band hemoglobin does, so no colour rule can reject it. Local contrast and your own eyes do.
• CLIPPED means the pixel is blown out and the reading is worthless. Move out of the sun.