Connecting Your Camera to the API

Method 1 · Recommended — zero server needed

Camera HTTP Event

Best for: Axis, Hikvision, Dahua, Milesight, and other cameras with HTTP event actions.

Many modern IP cameras can send an HTTP POST with a snapshot when they detect motion or a vehicle. Point that action directly at the API and you have plate recognition with no server, no middleware, and no code.

1. 01 Open your camera's web interface.
2. 02 Go to Events → Motion Detection or Vehicle Detection (naming varies by manufacturer).
3. 03 Set the action to HTTP Notification or Upload to HTTP Server.
4. 04 Configure the POST target as shown below, attaching the snapshot as multipart form-data in a field named "image".
5. 05 Save, drive a car through, and check the response log.

• Axis: Events → Recipients → add an HTTP recipient with the URL and Authorization header, then a rule that sends a snapshot on motion.
• Hikvision / Dahua: Event → Smart Event → Line Crossing or Intrusion Detection → set the upload method to HTTP with the endpoint below.
• Milesight: Event → HTTP Notification → configure the POST URL and token.

Method 2

RTSP Stream + a Small Script

Best for: Any camera with an RTSP stream — which is nearly every IP camera made in the last decade.

Run a lightweight script on any machine on the camera network — a Raspberry Pi, a spare PC, or an existing server. It grabs frames from the RTSP stream and posts them to the API. You get full control over capture frequency, motion gating, and what happens with each result.

import cv2
import requests
import time
import io

API_URL = "https://api.example.com/v1/detect"
API_KEY = "YOUR_API_KEY"
CAMERA_RTSP = "rtsp://admin:[email protected]:554/stream"

cap = cv2.VideoCapture(CAMERA_RTSP)

while True:
    ret, frame = cap.read()
    if not ret:
        continue

    # Encode frame as JPEG
    _, buffer = cv2.imencode('.jpg', frame)
    image_bytes = io.BytesIO(buffer.tobytes())

    # Send to the recognition API
    response = requests.post(
        API_URL,
        headers={"Authorization": f"Bearer {API_KEY}"},
        files={"image": ("frame.jpg", image_bytes, "image/jpeg")},
    )

    result = response.json()
    for plate in result.get("plates", []):
        print(f"Plate: {plate['text']} "
              f"({plate['confidence'] * 100:.1f}% confidence)")

    time.sleep(1)  # 1 frame per second is enough for parking

cap.release()

• — Don't send every frame — 1 – 2 frames per second is enough for parking scenarios.
• — Add motion detection (OpenCV BackgroundSubtractorMOG2) to send frames only when a vehicle is present. This keeps API usage proportional to traffic.
• — Set plate_only=true if you don't need car brand/model/color — it's roughly 2.5x faster.

Method 3

FTP Snapshot Folder

Best for: Cameras that save snapshots to FTP on motion or event triggers.

Many cameras can drop a JPEG onto an FTP server when they detect motion. A small folder-watcher script picks up each new file, posts it to the API, and moves it to a processed directory. Simple, robust, and easy to debug — every frame is a file you can inspect.

import os
import time
import requests

WATCH_DIR = "/path/to/ftp/snapshots"
PROCESSED_DIR = "/path/to/processed"
API_URL = "https://api.example.com/v1/detect"
API_KEY = "YOUR_API_KEY"

while True:
    for filename in os.listdir(WATCH_DIR):
        if filename.lower().endswith(('.jpg', '.jpeg', '.png')):
            filepath = os.path.join(WATCH_DIR, filename)

            with open(filepath, 'rb') as f:
                response = requests.post(
                    API_URL,
                    headers={"Authorization": f"Bearer {API_KEY}"},
                    files={"image": f},
                )

            print(f"{filename}: {response.json()}")

            # Move to processed folder
            os.rename(filepath, os.path.join(PROCESSED_DIR, filename))

    time.sleep(2)