Drone Detection System

As part of a symposium presentation, I decided to build several drone detection prototypes using different approaches: image-based detection, audio, and radio frequency.

For image-based detection, my initial idea was to train a machine learning model and deploy it on a dedicated Linux board with a camera array. Instead, I chose a simpler and more engaging approach that attendees could easily replicate—even without much maker experience.

SenseCAP Watcher

SenseCAP Watcher is an AI observer developed by Seeed Studio to monitor anomalies within a space and trigger actions. It includes a camera, touchscreen, speaker, microphone, WiFi, Bluetooth, RGB LED, I2C, UART, and a rotary knob with a button.

It connects via WiFi to an app called SenseCraft, where the device can be configured.

So how do you make SenseCAP Watcher detect a drone?

• Open the app
• Pair the device and select SenseCAP Watcher (the app supports multiple devices)
• Describe, in natural language, what to detect and how to react. Example: “If you see a drone, notify”

SenseCAP creates the task and enables a Details Config button. There you can choose whether to use a TinyML model or a Vision LLM. You can train a custom model for non-standard detections or rely on the Vision LLM for common objects. Since drones are a standard object, no additional ML work was required.

You can then define how the device reacts: LED flash, on-screen message with sound, push notification to the app, HTTP notification, or even UART serial output. You can also configure capture frequency and task duration.

Wio Terminal

SenseCAP supports multiple interconnection options, including serial communication. I connected it to another Seeed Studio device: the Wio Terminal, which I had previously used for a MasterMind game project. The Wio Terminal is based on a SAMD51 and includes a screen, buttons, buzzer, light sensor, IR, microSD, and GPIO.

I connected both devices using male-to-male jumper wires so that powering the Wio Terminal also powers the SenseCAP. I added two additional wires for UART serial communication.

Circuit

GND → GND

5V → 5V

Watcher pin 19 → Wio pin 10

Watcher pin 20 → Wio pin 8

Enclosure

Using Fusion 360, I designed an enclosure to house both boards, with a base compatible with a standard photography tripod mount.

Software

In Arduino IDE, add the board manager URL:

https://files.seeedstudio.com/arduino/package_seeeduino_boards_index.json

Install libraries:

#include <Seeed_Arduino_FS.h>
#include <ArduinoJson.h>

Select Seeeduino Wio Terminal, connect with a USBC cable and upload the code to Wio Terminal.

Usage

Operation is straightforward. The screen displays a log of detected drone events. Using the right button, you can scroll through the log, and pressing the button clears it. All detections are stored on the microSD card at:

/drone_log.txt

Demo

References

https://wiki.seeedstudio.com/sensecraft-ai/tutorials/sensecraft-ai-pretrained-models-for-watcher/

https://wiki.seeedstudio.com/Wio-Terminal-Getting-Started/