Kathmandu University

Department of Computer Science and Engineering

Dhulikhel, Kavre

A Project Proposal
On
Ocular Parking System
[Code No: ENGG 102]
(For the partial fulfillment of 1st Year/ 2nd Semester in Computer Science)

Submitted by:
Utsav Maskey (28)
Saugat Adhikari (03)
Suman Pokhrel (37)
Saharsa Ojha (30)

Submitted to:
Ms. Deni Shahi
Department of Computer Science and Engineering

Submission Date: 19th April, 2019

Abstract

Recent years have witnesses a growing interest in developing automatic parking

systems. However the efficient methods for locating parking slots using vision-based
system is still an under-developed area in Computer Vision. Ocular Parking System
is an Image Processing / Computer Vision project that intends to automate miniature
real-life tasks using computational techniques and resources. Given a parking lot
camera footage or image, the code turns it into a grid showing empty parking spaces.

This project aims to get us familiar with open source libraries and improve our
understandings on object-oriented as well as procedural programming concepts. We
like to settle the problem by methods of acquiring, processing, analyzing and
understanding digital images from real world to produce symbolical information. We
specifically focus on the use of open-source C++ libraries like OpenCV for image
processing tasks. Our approach is likely to benefit institutions, offices, etc. with pre-
installed CCTV cameras in parking lot. It can be further fine-tuned to solve other
similar tasks like: empty seats detection in a hall.

The project will be experimental and we expect it to be error-prone. But its use is
substantial considering its minimal cost and its usage with already accessible
resources.

Keywords: Image Processing, Computer Vision, Parking Assistant Systems

II
Table of Contents:

Title Page No.

Abstract ………………………………………………….........…........... II

Acronyms …………………………………………………………………. IV

Chapter 1: Introduction ………………………………………………….. 1-2

1.1 Background ………………………………………………………... 1

1.2 Objectives ……………………………………………………….. 2

1.3 Motivation and Significance ………………………………...……...... 2

Chapter 2: Related Works/Existing Works ………………….……….. 3

Chapter 3: Procedure and Methods …………………………..………. 4-7

Chapter 4: System Requirement Specification …..…………………... 8

4.1 Software Specification ……………………………..………….….. 8

4.2 Hardware Specification ………………………...………………...... 8

4.3 Optional Requirements ………………………...………………...... 8

Chapter 5: Project Planning and Scheduling ………...……………….. 9

References ………………………..…………………………………….... 10

III
List of Figure

Figure Page No.

Fig 3.1: Flow diagram of the main program …………………………………… 5

Fig 3.2: Flow of function “transformation”…………………………………….. 6
Fig 3.3: Flow of function “locate_grid” ….…………………………………….. 6
Fig 3.4: Flow of function “locate_vehicles”……………………………………. 6

List of Tables

Table Page No.

Table 5.1: Gantt Chart………………………………………………………… 9

Acronyms

CV Computer Vision
OS Operating System
RAM Random Access Memory
PAS Parking Assistant Systems

IV
Chapter 1: Introduction

1.1: Background

As years goes by, image recognition technology has been growing rapidly. As the
market has been saturated with cheaper, smaller and higher quality camera, the
advanced and computational technology has gradually become an essential part of
everyday life. The growth of number of vehicles in a city leads to issue of limited
parking spots available, with the help of CV, it has become easier and faster for
creating a special monitoring system which searches for free parking space in order to
reduce the amount of time required for searching the vacant space. The common
method that is used for the monitoring of parking lot is by using different type of
sensors. However, this causes the problem of motion detection and tracking. So, we
can eradicate this problem by using CV.

As previously stated, PAS generally use network of sensors which includes weigh-in-
motion sensors which are embedded into the parking space. But, this process is not
cost effective for large parking spots because individual sensors can cost a lot of
money. In comparison to sensor based parking detection, which requires installation
and maintenance of network of sensor, vision based systems are more cost effective
and easy in maintenance. So, instead of using network of sensors, we can create a
system using OpenCV library in C++ in order to monitor the parking lot and detect
empty spaces. This can be made possible by continuous experimentation until the
desired output is obtained.

1
1.2. Objectives

 To detect empty spaces in parking lot.

 To benefit small scale organizations.
 To bridge the gap between human vision and that of computers.
 To further extend our project to meet market’s requirements.

1.3 Motivation and Significance

We find that in the field of vision-based parking slot detection, there is still large
room for further improvements. The first motivation to study computer vision is to
gain an insight into the fundamentals of information processing in vision. The rise of
image-processing tools and methodologies further motivates automation of real-life
vision tasks.

2
Chapter 2: Related Works/Existing Works

We focus on studying the problem of in-vehicle vision-based parking-slot detection.

Some of the representative works in this area are mentioned here.

Various methods for the monitoring of parking lot includes different type of sensors
like: overhead indicator sensors, surface mount sensors, in-ground sensors, ultrasonic
sensors, etc which are cost inefficient.

Moreover, numerous image processing methodologies have also been tested.

According to (Jan, 2016), the problem can be broken down into two steps:
determination of parking spots from image and to determine whether or not there is a
movement on the parking lot. Considering vehicles can be parked for longer hours,
the movement determination might not be a proper solution.

Machine-learning techniques have also been implemented. (Adam, 2019), proposes

that the regions of image with non-moving cars can be assumed to be parking spaces.
But the use of deep-learning techniques to detect vehicles and carrying out real-time
model training can be computationally expensive.

We extract ideas and concepts from these works as well and address the program to
be applicable as well as cost-efficient.

3
Chapter 3: Procedure and Methods

The methodologies that we plan to use for the completion of the project are listed
below.

1) Study and research: Vehicle detection system is a challenging problem. So

firstly, we planned to do some research related to the tools that we will be
using. We plan on studying about the tools and gain some ideas with the help
of online courses, tutorials, reference books and research papers. This helps us
develop a clear sense of direction early on in the project and also help us
organize our project.

2) Core Programming: In this part we plan to implement the ideas gathered

from our research and study, in our program using the available tools. We
heavily focus the use of open-source library, OpenCV. In this section we plan
to analyze the images (preferably video footage) until the desired output is
obtained.

3) Evaluation and Parameter tuning: The developed program is sure to be

error-prone. So, we will fine tune our parameters for better accuracy. The
study generalizes our understanding and this part particularizes our idea.

4) Documentation: After the above mentioned tasks are completed we will

prepare work report and present our project.

4
We put forward the following flow diagram:

Fig 3.1: Flow diagram of main program

5
Fig 3.2: Flow of transformation Fig 3.3: Flow of locate_grid

Fig 3.4: Flow of locate_vehicles

6
Steps:

1. Perform initial transformations

We take in two input images: empty parking lot image “img_empty” and test-
time image containing vehicles “img_test”. The images are first gray-scaled to reduce
complexity. Then edges of images are detected using appropriately tuned Sobel-filters.
Finally we perform Binary Thresholding operations to facilitate Binary operations in
images.

2. Locate the co-ordinates of parking grids and vehicles

The image “img_empty” is used to detect the parking grids. The thinning
operations are performed on gray-scaled images. Then we find the co-ordinates of
end-points and branching points using techniques like template matching or Harris
corner detection.

3. Create a symbolical figure showing parking grids and empty

spaces
We output the information in symbolical format, most preferably the parking
grids with rectangles. The regions containing vehicles are marked with red rectangles
whereas green rectangles mark the empty spaces.

7
Chapter 4: System Requirement Specification

4.1 Software Specification

4.1.1 C++ dependencies: OpenCV >= 3.0 (tested on version 3.2.0)
4.1.2 OS: Linux, Mac, Windows (64-bit Recommended)

4.2 Hardware Specification

4.2.1 CPU: 64-bit based processor (Modern CPU recommended for real-
time processing)
4.2.2 RAM: About 2GB of RAM Recommended

4.3 Optional requirements

4.3.1 Camera: Any camera supported by OS (only for video footage input)

8
 Chapter 5: Project Planning and Scheduling
The work breakdown and time in weeks required to complete the specific task are
shown as in the Gantt chart below: -

Task (Weeks) 1 2 3 4 5 6 7 8 9 10 11 12

Research and Study

Core Programming

Evaluation

Parameter Tuning

Documentation

Table 5.1: Gantt Chart

Tasks:
1. Research and Study
2. Core Programming
3. Evaluation
4. Parameter Tuning
5. Documentation

9
References
[1] Baggio, D. L., Escrivá, D. M., Emani S. (2017). Mastering OpenCV 3: Second
Edition - Get hands-on with practical Computer Vision using OpenCV 3

[2] Laganière, R. (2017), OpenCV 3 Computer Vision Application Programming

Cookbook

[3] Object Oriented Programming (2014)

Online Course - edx.org

[4] Image and Video Analysis (2015)

Youtube video tutorials by pythonprogramming.net

10