Department Logo
Team #01  |  ECE 196  |  Spring 2025 – UC San Diego

Team Members

zlf

Linfeng Zhang

wsh

Sihan Wang

lyh

Yu-Heng Lin

Team's Problem: Micro-mobility Device Theft on Campus

Micro-mobility device theft is a common issue at UC San Diego, where limited tracking and monitoring make it difficult to prevent or resolve stolen e-scooters and bikes.

Problem Motivation

  1. We chose this issue because many students rely on personal micro-mobility devices for their daily commutes across campus. Nobody wants to walk uphill both ways with a midterm in 10 minutes :(
  2. The growing number of thefts has created financial stress, safety concerns, and student transportation disruptions. It's a problem affecting the quality of life and trust in campus safety services.
  3. Maybe the owner has special connection to their micro-mobility devices. We want to prevent people from losing their beloving ones.

Current Existing Solutions

GPS Tags & Trackers
  • Use: Paired with an app to track and trigger alerts.
  • Pros: Real-time tracking; recovery aid.
  • Cons: Can be disabled; requires power and signal.
GPS Trackers
U-locks & Chains
  • Use: Lock to bike racks or fixed objects.
  • Pros: Easy to use and common.
  • Cons: Can be cut; offers no alerts.
Locks

Our Proposed Solution

Block Diagram
Block Diagram
Technology:
  • Detect if the micro-mobility device is moved or stolen using accelerometer
  • Trigger an audible alarm in real-time when theft is detected
  • Track the live GPS location of the stolen device
  • Capture and upload a photo of the thief using a built-in camera
  • Detect the device owner using face recognition
  • Allow owners to monitor and control the system through a web interface
Technological Aspects of Solution
  • ESP32 microcontroller with built-in WiFi for communication
  • Accelerometer and gyroscope module for motion detection
  • GPS module for real-time tracking
  • Camera module (ESP32-CAM) for facial image capture
  • Machine learning-based face recognition using a ResNet-34 Convolutional Neural Network (CNN) to generate 128-dimensional facial embeddings for identity verification
  • Web server hosted on the cloud interacting with ESP32 microcontroller for real-time monitoring
  • Alarm module to alert nearby people

Hardware Design Overview

PCB Schematic
PCB Schematic
3D PCB View
3D PCB View
PCB Purpose

To develop a smart security system for devices that detects unauthorized movement, sends real-time location updates, and provides alerts using a buzzer, with an integrated LED and switch for power control.

ESP32-CAM & Main PCB Enclosure CAD
ESP32-CAM Enclosure CAD
ESP32-CAM Enclosure
System Operation Flowchart
System Flowchart
Why Face Recognition?

Instead of capturing every image when motion is detected, which leads to storing lots of unnecessary photos. We now use a smarter, more efficient approach.

Our upgraded system uses face recognition, which works in two steps:

  1. Step 1: Detect whether a human face is present.
  2. Step 2: Compare that face with the owner's or their registered friends'.

Only relevant images are saved, such as an unrecognized person tampering with the device. This helps conserve storage and improves security.

Users can also upload their friends' faces so trusted people can move or retrieve the scooter without triggering alerts.

Milestones

Project Timeline
  • 📅 By Sunday of Week 6: ✅ Done
    Complete the web server, including login, register, profile, and face recognition pages. Test and ensure Wi-Fi functionality on the ESP32 development board.
  • 📅 By Sunday of Week 7: ✅ Done
    Finish the main PCB board (with accelerometer, GPS, alarm, and ESP32). Deploy the web interface to allow communication with the ESP32 microcontroller.
  • 📅 By Sunday of Week 8: ✅ Done
    Read movement and GPS data from the microcontroller, trigger the alarm, and solder the ESP32 board. Integrate the ESP32-CAM camera into the system.
  • 📅 By Thursday of Week 9: ✅ Done
    Connect image capture from the camera to the web application and complete the face detection functionality.
  • 📅 By Monday of Week 10: ✅ Done
    Model and 3D print the final enclosure for the hardware.

  • Progress Update:

    • Completed computer vision module (face detection & recognition)
    • Built web interface (login, register, profile)
    • Successfully tested Wi-Fi connectivity using an alternate ESP32 board
    • Assembled and test the main PCB with every component being functional
    • Wrote the firmware for buzzer, gps, and accelerometer
    • Read GPS and movement data from the microcontroller, integrated the ESP32-CAM camera into the system
    • Sent the movement and GPS data from the microcontroller to the web, and sent images/streaming from ESP32 Cam to the web
    • Designed the CAD Model by Onshape and 3D printed the final enclosure for the hardware

    • Resources

    We'd like to acknowledge the following resources that supported our project development.

    • 🔗 ESP32 with MPU-6050: Tutorial by Random Nerd Tutorials
      Helped us set up and wire the accelerometer (MPU-6050) correctly on our main PCB.
    • 🔗 ESP32 with NEO-6M GPS Module: GPS Setup Guide
      Guided our GPS wiring and use of UART communication protocol as recommended.
    • 🔗 WiFi Client with FreeRTOS: GitHub Repository
      We used their example code to establish a stable WiFi HTTP client on our ESP32 board.
    • 🔗 ESP32-CAM Video Streaming Server: ESP32-CAM Streaming Guide
      We adapted the camera server implementation to enable bidirectional HTTP functionality.