Charlot DEDJINOU
Back to projectsFlagship case study · 2025
  • AIoT
  • Computer vision
  • Embedded systems

Smart mobility · Benin

Smart ParkingAIoT System

A four-space prototype connecting mobile reservation, secure physical access, camera-based occupancy and live operator supervision.

The model joins four marked spaces, an elevated camera, local feedback and a motorised barrier in one working scale environment.
04
Parking spaces
10 s
Capture interval
04
Connected layers
01
Integrated prototype

First Prize

Smart Cities Hackathon · FRIARE · April 2025
Recognition / 2025

01 · The problem

Urban growth is outpacing the way parking is managed.

In Benin's growing urban areas, parking infrastructure and management practices struggle to keep pace with demand. Informal roadside parking reduces usable public space, disrupts traffic and exposes drivers and pedestrians to avoidable risk.

Existing local services provide only parts of the experience. Drivers still lack a reliable way to reserve before travelling, identify an available space on arrival and enter without manual intervention. Operators lack an integrated view of occupancy and parking activity.

Pressure

Growing urban demand

More movement and economic activity place increasing pressure on limited parking space.

Friction

Disorganised parking

Roadside and informal parking obstruct traffic, consume public space and increase safety risks.

Gap

Limited local automation

Reservation, guidance, physical access and supervision rarely operate as one locally adapted system.

The guiding question

How can a technologically advanced solution, adapted to the local context, improve parking management in Benin’s urban areas?

02 · Mobile application

Reserve before arriving.

The mobile app turns live availability into a guided booking flow and carries the same reservation through payment and physical access.
  1. 01

    Check availability

    Free spaces and occupancy update on the home screen.

  2. 02

    Choose a slot

    Arrival time, duration and accessibility needs shape the request.

  3. 03

    Confirm and pay

    A deposit confirms the booking and unlocks the access workflow.

  4. 04

    Retrieve access

    The reservation remains available with its status and access code.

03 · End-to-end architecture

One operational loop. Four product layers.

The mobile app initiates user actions. FastAPI applies business rules and synchronises state. The AIoT core observes the physical parking. The dashboard turns the shared state into operator visibility.

01
MobileReservation, payment and access code.
02
BackendValidation, APIs and WebSocket updates.
03
AIoT coreCapture, vision and occupancy decisions.
04
DashboardMonitoring and operations.

The physical prototype

Software decisions reach the gate.

The model joins four marked spaces, an elevated camera, local feedback and a motorised barrier in one working scale environment.

  • ESP32
  • ESP32-CAM
  • 4×4 keypad
  • LCD 16×2
  • Servo gates
  • Ultrasonic sensing
Keypad, LCD, ESP32 controller and barrier actuation on the access prototype.

04 · Vision pipeline

Turning one camera frame into four parking states.

Each parking ROI is corrected for perspective before YOLO measures detected vehicle area. A 2% ratio threshold updates the slot state.

A custom PyQt editor records the geometry of each space from the real camera view.
Qualitative checksObserved scenarios · no accuracy claim
Scenario A · All four spaces observed as occupied.
Scenario B · Two occupied spaces and two free spaces.

05 · Access control

A reservation opens a real gate.

The access code connects a digital booking to a physical action while local hardware keeps the interaction readable.

  1. 01

    Enter the six-digit code

    The keypad captures the access credential and allows correction before submission.

  2. 02

    Validate through FastAPI

    The backend checks the reservation, payment status and time window.

  3. 03

    Guide and open

    The LCD displays feedback and a free-space number before the servo lifts the barrier.

  4. 04

    Detect passage and close

    The ultrasonic signal helps the controller close the gate after the vehicle clears the entrance.

06 · Operator supervision

The same parking state, made operational.

The dashboard brings availability, occupancy, active reservations and administrative data into one supervisory interface.

  • Live availability
  • Reservations
  • Payments
  • Users

07 · Engineering decisions

The difficult parts shaped better tools.

Calibration

Manual ROI definition was too slow.

A dedicated PyQt editor made the slot geometry visual, adjustable and persistable.

Concurrency

Entry and exit had to work together.

Two FreeRTOS tasks run on the ESP32 cores while a mutex protects shared Wi-Fi and HTTP access.

Electronics

Voltage and pin constraints changed the wiring.

A suitable 5 V supply and cascaded I2C expanders supported the LCD and keypad without exhausting ESP32 pins.

What is demonstrated

  1. 01Reservation and deposit workflow in the mobile application.
  2. 02Six-digit access validation and physical gate actuation.
  3. 03ESP32-CAM acquisition and qualitative occupancy decisions.
  4. 04Shared backend state exposed to user and operator interfaces.

What remains to validate

  1. 01Precision, recall, mAP and decision-level F1.
  2. 02Robustness under low light and camera pose drift.
  3. 03Offline resilience and later state synchronisation.
  4. 04Deployment across larger sites and multiple cameras.

The outcome

A working AIoT prototype where product software, computer vision and embedded hardware share one operational state.

The four-space model proves the complete interaction. Quantitative vision validation and site-scale resilience define the next engineering stage.