Halo
RM.
Computer Vision

Unauthorized Truck Detection via CCTV

Computer-vision security system that reads truck plates from CCTV streams, checks a whitelist, and sends WhatsApp/Telegram alerts.

Role
Computer Vision Engineer
Year
2026
Technologies
PythonYOLO (Ultralytics)FastAPIPyTorch
UT

Overview

An automated system that monitors and flags unregistered trucks entering a restricted area in real time, without a manual operator.

The Problem

Vehicle access control in restricted areas usually relies on staff watching CCTV — slow and easy to miss.

Approach

RTSP streams go through two stages: YOLO detects vehicles crossing a virtual line, an IoU tracker follows them, a second YOLO finds the plate, then OCR reads it and a rule engine checks it against the whitelist.

Solution

A FastAPI backend for event ingestion, whitelist sync and camera management; instant alerts through Twilio WhatsApp and Telegram; a YOLOv8/YOLO11 training pipeline for custom vehicle datasets.

Model Details

Problem
Vehicle & licence-plate detection, then plate OCR from CCTV video.
Model
YOLOv8/YOLO11 (vehicle & plate) + PaddleOCR
Deployment
RTSP stream inference + FastAPI REST API