Multi-Sensor Human Detection System

Demo Video

A real-time embedded system using three LD2410S mmWave radar sensors connected to a Raspberry Pi 4B. The system detects human presence, estimates distance, and determines direction using multi-sensor fusion, with live visualization via a web dashboard.

Designed as a real-world embedded system integrating hardware sensing, low-level communication, and networked visualization on a single-board computer.

git clone <repo>
cd ld2410-human-detection
bash setup.sh
python3 src/main.py

Open: http://<PI_IP>:8000

This project demonstrates:

• Real-world embedded system design
• Multi-UART hardware interfacing on Raspberry Pi
• Low-level binary protocol handling (no third-party sensor libraries)
• System-level debugging (power supply, UART config, wiring)
• Integration of hardware, software, and networking

• Real-time multi-UART sensor acquisition (3 sensors simultaneously)
• Custom binary protocol parser — no external libraries
• Sensor fusion for LEFT / CENTER / RIGHT direction estimation
• Fault-tolerant UART handling with automatic reconnect
• Live monitoring via web dashboard (accessible from any device on the same network)

Live dashboard served at http://<PI_IP>:8000 — auto-refreshes every 500 ms. Shows presence, distance, and direction for all three sensors.

Always cross TX/RX: Sensor OT1 (TX) → Pi RX pin, Sensor RX ← Pi TX pin.

See docs/hardware_setup.md for full wiring table.

git clone https://github.com/<your-username>/ld2410-human-detection.git
cd ld2410-human-detection

sudo nano /boot/config.txt

Add at the bottom:

enable_uart=1
dtoverlay=disable-bt
dtoverlay=uart3,txd3_pin=4,rxd3_pin=5
dtoverlay=uart4,txd4_pin=12,rxd4_pin=13

Save and reboot:

sudo reboot

sudo raspi-config
# Interface Options → Serial Port
# Login shell: No  |  Hardware enabled: Yes

bash setup.sh
# OR manually:
sudo apt install python3-serial
sudo usermod -aG dialout $USER

Log out and back in (or reboot) for the group change to apply.

ls /dev/ttyAMA*
# Expected: /dev/ttyAMA0  /dev/ttyAMA3  /dev/ttyAMA4

Run the hardware test:

python3 test_uart.py

python3 src/main.py

Open in browser: http://<PI_IP>:8000
Find your IP: hostname -I

ld2410-human-detection/
│
├── src/
│   ├── main.py           # Entry point, threads, direction logic
│   ├── uart_reader.py    # Serial port open/read/reconnect
│   ├── sensor_parser.py  # LD2410S binary frame decoder
│   └── web_server.py     # HTTP server (dashboard + JSON endpoint)
│
├── docs/
│   ├── architecture.svg         # System architecture diagram
│   ├── dashboard_screenshot.svg # Web dashboard preview
│   ├── hardware_setup.md        # Wiring diagrams, component list
│   ├── uart_config.md           # /boot/config.txt guide
│   └── troubleshooting.md       # Common issues and fixes
│
├── test_uart.py     # Hardware verification script (run first)
├── setup.sh         # One-time setup script
├── requirements.txt # Python dependencies
└── README.md

See docs/troubleshooting.md for detailed solutions.

• GPS tagging — attach coordinates to detection events (u-blox NEO-6M)
• Camera integration — visual confirmation with OpenCV + Pi Camera
• MQTT / IoT dashboard — publish to Node-RED or Grafana
• AI presence classification — TensorFlow Lite on-device inference
• Systemd service — auto-start on boot
• Alert system — SMS/email on detection via Twilio or SMTP

Gagan Manjunath
NIE - South, Mysuru Electronics and Communication Engineering (ECE)

Srujan H R Yuvaraja's College, Mysuru BSc (Maths, Statistics, Computer Science)