School of Electronics and Telecommunications: Thesis

You might also like

Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 7

HANOI UNIVERSITY OF SCIENCE AND TECHNOLOGY

SCHOOL OF ELECTRONICS AND TELECOMMUNICATIONS

THESIS

SUBJECT: TECHNICAL WRITING AND PRESENTATION


TOPIC: PARKING LOTS DETECTION USING PROCESSING IMAGE

INSTRUCTOR: Nguyen Tien Hoa


STUDENT: Nguyen Hong Quang
STUDENT ID: 20182738
CLASS: 07-K63
Hanoi, November 2020

Table of Contents
LIST OF FIGURES 2
ABSTRACT 3
CHAPTER 1. INTRODUCTION 4
1.1. Information 4
1.2. The equation 5
1.3. The result of project 5
LIST OF REFERENCES 6

1
LIST OF FIGURES
Fig 2.1: The result 5

2
ABSTRACT
There are a lot of methods in order to detect and track parking lots in the world. The
objective of this plan is to detect vehicles in parking spaces by a basic method which
everybody can understand and do not take too much time. Accordingly, the method when
detecting and tracking vehicles is Background Subtraction. The core of this work is to
evaluate the difference between background image and parking lot image using function in
the OpenCV.1 Two techniques are combined: background subtraction using Gaussian
Mixture and ‘vehicles’ contours drawing. The method can apply a few real-life scenarios
while being a low computational cost solution which can be applied to kind of parking areas
within a camera’s reach.

1
Dr. Adrian Rosebrock. Practical Python and OpenCV: An Introductory, Example Driven Guide to
Image Processing and Computer Vision.

3
CHAPER 1. INTRODUCTION
1.1. Information
       Nowadays, in the world, the use of smart parking systems increases quickly and is an
important part of the society. This system can be helpful with drivers when they want to find
a place in order to park while yet do not know whether there is an available parking lot or
not. Moreover, smart parking systems will send notifications to the manager and then they
can resolve such issues as: the vehicle parking on more than a slot, objects lying on where
the vehicle can park, etc…
       One of the methods is processing images using deep learning. As computers can
study quickly, they can process data faster with higher accuracy. For example, David Yu 2
used AI and GIS in parking lot detection. In order to use this method, we can apply the
knowledge of AI, processing data using Drone2Map and performing data using TensorFlow
and ArcGIS Pro. We can also use YOLO 3 in machine learning in order to detect a place
where the vehicle parks. YOLO is a CNN network model for object detection, recognition,
and classification. YOLO is made up of convolutional layers and connected layers, in which
convolutional layer extract features of the image and full-connected layers predict the
probability and the coordinates of the object. The meaning of YOLO is “You only look
once”. This method has a high processing speed, which can detect objects just by one look,
so the manager needs to be trained carefully with computer vision. On the other hand, both
techniques are absolutely difficult for newbies. Users must have a good knowledge base
about AI and know how to use YOLO. Moreover, you must provide computers with a lot of
image data.
Many techniques have been proposed to address the parking lots classification
problem using computer vision techniques. Some of them are based on single-image
analysis. I chose it because this method is very easy; however, I have had a lot of difficulties
in image processing. For example, when a parking lot has a big object which is of the
comparatively same size with a vehicle, then my software detects it to be a vehicle.
Otherwise, shadows of trees or vehicles may dump in the parking lots and the software may

2
Yu, D., 2018. Parking Lot Vehicle Detection Using Deep Learning. [online] medium.com.
Available at: https://medium.com/geoai/parking-lot-vehicle-detection-using-deep-learning-
49597917bc4a
3
Hoàng, V., 2019. Tìm Hiểu Về YOLO Trong Bài Toán Real-Time Object Detection. [online] Viblo.
Available at: https://viblo.asia/p/tim-hieu-ve-yolo-trong-bai-toan-real-time-object-detection-
yMnKMdvr57P?fbclid=IwAR306SaL09R-uC09_jiwP-96pFrH08VwuLs0Ywgphix8NWSzN-
9QQ3VGNNc

4
mistake them for real vehicles. Because binary has two values 0 or 1, we cannot determine
whether there is a vehicle or not.

1.2. The equation


n m
SAD=∑ ∑ Ps ( Xi , Yj )−Pbg( Xi ,Yj )
i=0 j=0

We define:
● SAD: sum of absolute different
● Pbg: pixel of background
● Ps: pixel of parking lots

1.3. The result of project

Fig2.1: The result

5
LIST OF REFERENCES
1. Dr. Adrian Rosebrock. Practical Python and OpenCV: An Introductory, Example
Driven Guide to Image Processing and Computer Vision.

2. Yu, D., 2018. Parking Lot Vehicle Detection Using Deep Learning. [online]
medium.com. Available at: https://medium.com/geoai/parking-lot-vehicle-detection-
using-deep-learning-49597917bc4a

3. Hoàng, V., 2019. Tìm Hiểu Về YOLO Trong Bài Toán Real-Time Object Detection.
[online] Viblo. Available at: https://viblo.asia/p/tim-hieu-ve-yolo-trong-bai-toan-real-
time-object-detection-yMnKMdvr57P?fbclid=IwAR306SaL09R-uC09_jiwP-
96pFrH08VwuLs0Ywgphix8NWSzN-9QQ3VGNNc

4. Liu, S.: 'Robust vehicle detection from parking lot images'. Technical Report no.
ECE-2005-06, Boston University, Department of Electrical and Computer
Engineering, 2005, pp. 10-20.

5. Carlos Gálvez del Postigo , Juan Torres, Jose-Manuel Menéndez. Vacant parking
area estimation though background subtraction and transience map analysis.

You might also like