1 of 25

Implementation of an Intelligent Parking System

​

Dr. Jehangir Arshad Adnan Yousaf Ateeq Ur Rehman

Chaudhry Ahmed Ijaz Sania Habib Huzafa Abid

Rao Muhammad Asif

1

International Conference on Engineering and Emerging Technologies

Co-Authors:

2 of 25

Presentation Contents

  • System Overview
  • Methodology
  • Results
  • Conclusion

​

​

​

​

​

2

3 of 25

System Overview

  • Intelligent Parking System
  • Vehicle Identification at entrance
    • Number Plate Scanning
  • Exit System
    • Number plate scanning
    • Fare calculation according to duration of stay
  • Fare Records for Owner
    • Desktop Application

​

​

​

  • Real-time monitoring
    • Mobile app for drivers
  • Security features
    • Login Authentication for guard
    • Password on owner’s app

​

3

4 of 25

Proposed System

4

Overview of Parking lot

Overview of Backend System

5 of 25

Objectives�

    • Development of Cloud Database to store parking status
    • Development of backend image processing algorithm for free slot detection
    • Integration of backend image processing algorithm with the cloud database
    • Development of image processing algorithm for vehicle and license plate detection at entrance and exit
    • Development of fare calculation algorithm
    • Development of driver side apps for real-time monitoring
    • Development of owner/admin side desktop app for overall monitoring of system

​

5

6 of 25

Methodology at Entrance

  • Whenever an object is at entrance, there are two very most important steps to follow:
    1. Object Identification
      • Single Shot Detection (SSD) MobileNet deep learning model is used

​

​

​

​

​

​

6

Layer architecture of the SSD model

7 of 25

Methodology at Entrance

7

  • Performance analysis of model:
  • SSD MobileNet gives better balance of speed and accuracy on CPU than others

8 of 25

Methodology at Entrance

2. Object Tracking

      • Necessary to keep track or count of unique vehicles
      • Algorithm:

​

​

​

​

​

​

​

8

9 of 25

Applications

9

Splash Screen

Admin Application [Windows Form.Net]

Password Window

Dashboard

10 of 25

Applications

10

Admin Application

Real-Time Status

11 of 25

Applications

11

Admin Application

Fare Collection Window

12 of 25

Applications

12

Admin Application

Settings

13 of 25

Applications

13

Mobile Application [Xamarin.Net]

Splash Screen

Real-Time Status

14 of 25

Structure of Database [Firebase]

14

Database Structure Overview

15 of 25

Results At Entrance

15

Empty Entrance

Entrance with a person

16 of 25

Results At Entrance

16

1. Entrance with a vehicle

2. Entry in database

3. Owner Application

17 of 25

Methodology and Results of License plate Detection

17

1. Vehicle image from Entrance Stage

2. Cropping the image for better detection

3. Applying image processing’s morphological operation to get closed binary image

4. Finding and cropping contours having exactly four vertices and area greater than a calculated threshold

5. Sending the image to the *OCR server

6. OCR Output

*Google Vision API is used as online OCR server

18 of 25

Methodology and Results of Parking Area

18

1. Marking of slots to be considered in the system

2. Applying image processing’s adaptive thresholding to separate object and background

Count number of white pixels in the slots

19 of 25

Methodology and Results of Parking Area

19

3. If count below a *specific threshold, assign green color to the slot to indicate free slot.

4. Repeat step 2 on frame with vehicle. Number of white pixels increases in a slot.

20 of 25

Methodology and Results of Parking Area

20

5. If count greater than a *specific threshold, assign red color to the slot to indicate occupied status.

*Experimentally calculated threshold: 2400px

Results of the applications and database

21 of 25

Results At Exit

21

When a vehicle is in front of exit

Results of the application and database

22 of 25

Performance Analysis

22

Actual Number

Predicted by OCR Server

Accuracy [%]

Time Taken [ms]

LEC-14 4555

LEC-S 4555

0

508

LEE-12 4769

LEE-12 4769

100

622

AMV-796

AMV-796

100

507

AJP 783

AJP 783

100

518

LE-16 5912

LE-16 5912

100

528

LEF-16 283

LEF-16 283

100

523

LEA-16 3144 A

LEA-16 3144 A

100

588

LEA-15 1540

LEA-15 1540

100

556

LEB-06 5700

LEB-06 5700

100

561

LEC-11 3378

LEC-11 3378

100

522

OCR Performance

It was observed that internet speed of 8MBps, it takes an OCR server an average time of 543.3ms to provide a result with 90% accuracy.

23 of 25

Performance Analysis

23

Cloud Database Performance

Time taken by our image processing system to save data into our cloud database. It is averaged to be 323.6ms on the internet speed of 8MBps.

Entry Record to be Saved

Time Taken [ms]

LEC-14 4555

312

LEE-12 4769

300

AMV-796

290

AJP 783

294

LE-16 5912

299

LEF-16 283

285

LEA-16 3144 A

366

LEA-15 1540

385

LEB-06 5700

364

LEC-11 3378

341

24 of 25

Conclusion

  • We are providing a cost effective solution that works on CPU instead of GPU
  • Fully autonomous entrance with semi-autonomous exit system having a reasonable processing time
  • Limitations:
    • Dependency upon light because of image processing
    • Image processing Issue when vehicle has similar color with background
  • Future Work:
    • Integration of EV Charging System
    • Integration of Booking System

​

24

25 of 25

​

​

​

​

Thank You

25