Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Xander44-4/traffic_reducer/llms.txt

Use this file to discover all available pages before exploring further.

Detection zones are the quadrilateral regions in the camera frame that correspond to each directional lane at the intersection. Every vehicle detected by YOLOv8 is assigned to one of four zones — Norte, Sur, Este, or Oeste — based on where its bounding-box centroid falls. Vehicles whose centroids do not land inside any zone are labelled other and excluded from the traffic counts used for signal-phase decisions.

​
Default zone definitions

Zones are defined in TrafficCamera.__init__() inside video_processor.py as normalized coordinate arrays. Each value is in the range 0.0–1.0 relative to the frame dimensions, so the same zone definition works regardless of the actual pixel resolution of the video feed. 
self.zones = {
    'norte': np.array([[0.00, 0.00], [0.43, 0.05], [0.43, 0.58], [0.27, 0.65]]),
    'este':  np.array([[0.47, 0.07], [1.00, 0.17], [1.00, 0.40], [0.57, 0.55]]),
    'sur':   np.array([[0.53, 0.55], [0.96, 0.42], [0.99, 0.94], [0.63, 1.00]]),
    'oeste': np.array([[0.00, 0.73], [0.55, 0.70], [0.43, 1.00], [0.00, 1.00]]),
}
These coordinates are tuned for the default 854 × 480 drone-view video (traffic_dron_view.mp4) and the bundled YouTube live stream. For a different camera angle, mounting height, or intersection geometry, the coordinates must be updated.

​
Zone color overlays

The live feed renders a colored polygon border around each zone to aid visual debugging. The overlay colors are also used for bounding boxes drawn around vehicles that have been matched to a zone.
ZoneOverlay colorBGR value
NorteAmber(255, 180, 60)
SurCyan(60, 200, 255)
EsteGreen(90, 230, 130)
OestePurple(200, 120, 255)
Unmatched vehicles (other) are drawn with a neutral grey box (130, 130, 130).

​
How to tune zones for a new camera

1

Capture a still frame

Pause the live feed or extract a single frame from your camera source. Save it as a PNG or JPEG so you can open it in an image editor.
2

Identify lane corners

In the image editor, locate the four pixel corners that bound each directional lane as seen from the camera’s perspective. Note down the (x, y) pixel coordinates for each corner.
3

Normalize the coordinates

Convert pixel coordinates to the 0.0–1.0 range using the frame dimensions:
x_norm = pixel_x / frame_width
y_norm = pixel_y / frame_height
For example, a point at pixel (427, 230) in an 854 × 480 frame normalizes to (0.50, 0.48).
4

Update the zones dictionary

Open video_processor.py and replace the coordinate arrays in TrafficCamera.__init__() with your new values:
self.zones = {
    'norte': np.array([[x1, y1], [x2, y2], [x3, y3], [x4, y4]]),
    'este':  np.array([[x1, y1], [x2, y2], [x3, y3], [x4, y4]]),
    'sur':   np.array([[x1, y1], [x2, y2], [x3, y3], [x4, y4]]),
    'oeste': np.array([[x1, y1], [x2, y2], [x3, y3], [x4, y4]]),
}
Restart the server and confirm the colored zone outlines align with the actual lane boundaries on the live feed.

​
Vehicle assignment logic

Zone matching is performed inside _run_yolo() using OpenCV’s cv2.pointPolygonTest(). For each detected vehicle, the method computes the centroid of its bounding box and tests it against each zone polygon in order: 
for zone_name, poly in self.zones.items():
    if cv2.pointPolygonTest(
        (poly * [w, h]).astype(np.int32), (cx, cy), False
    ) >= 0:
        counts[zone_name] += 1
        matched_zone = zone_name
        break
A return value ≥ 0 from pointPolygonTest means the centroid is on the boundary or inside the polygon. The loop breaks on the first match, so each vehicle is counted in at most one zone.
Zones can overlap. A vehicle centroid that falls inside an overlapping region is assigned to whichever zone is tested first. The fixed iteration order is Norte → Este → Sur → Oeste, matching Python’s dict insertion order as defined in __init__(). Design your zones to minimize overlap, particularly near the centre of the intersection.

Build docs developers (and LLMs) love

Get started for freeTalk to us