AI based Smart Parking System

2023 2nd International Conference on Advancements in Electrical, Electronics, Communication, Computing and Automation (ICAECA) | 979-8-3503-0681-1/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICAECA56562.2023.10200147

Ms. Bharathi Priya C Vishal KB Rajendiran Shivaramakrishnan

Computer Science Engineering Computer Science Engineering Computer Science Engineering
Kumaraguru College of Technology Kumaraguru College of Technology Kumaraguru College of Technology
Coimbatore, India Coimbatore, India Coimbatore, India
[email protected] [email protected] [email protected]
Devasenan Murugan
Computer Science Engineering:
Kumaraguru College of Technology
Coimbatore, India
[email protected]

Abstract— A new venture has been endorsed to thrive a user- a "Smart Parking System". The ideology behind Smart
friendly smart parking system with minimal human Parking System has always been in the government's future
interference. The proposed system scans for the vehicle type and plans and it is planned to be implemented soon in some
vehicle number when a new vehicle enters with the help of a important cities.
transfer-learning based model and a CNN model. A suitable
A unique innovation has been overtured on the account of the
parking spot is searched based on the vehicle type and assigned
to the vehicle. A customer would be monitored with the help of parking system, momentously cars in specific, which
webcams and number plate recognition model after entering the involves automatic allocation of parking spots and
parking lot and would be alerted through message if he/she has calculation of parking charges. The allocation of parking
parked his/her vehicle in a spot that is not assigned to him. The spots is based on the vehicle type and is managed through a
customer would then be displayed with a bill at the exit point website designed for parking lot owners. This proposal aims
with automated calculation of parking charges by the system. to upgrade the existing manual parking system to a smarter
YOLOv3 and Tiny-Yolov3 models has been proposed for vehicle and more efficient one, allowing for the number of vehicles
detection. The models have been compared by detection time in terms of greater volumes in a small space. The prime intent
per frame, detection accuracy and its F1 Score. On exploration,
of this Smart Parking System is to park vehicles in an
an EfficientNet based TL model, mobileNet Extended model
and light-weight CNN model has been proposed for vehicle type organized way and to provide a parking spot of suitable size
classification. Testing is done to compare the classification based on the vehicle type in an automated manner. This also
accuracy and time. The EfficientNet based TL model performed helps in the implementation of the smart city plan which is
extremely well in test data as well as in real-time data. For the future and which has been in plans of the government.
number plate detection, a pre-trained cascade classifier has been The Paper is structured as follows: Section I, outlines the
proposed. On the other hand, a CNN model, a ResNet50 Introduction, the detailed literature study is presented in
extended model and a ResNet152 extended model has been section II, Existing and proposed system, module description
proposed for the purpose of character recognition. The models is given in section III. The experimental results on various
have been compared based on their classification accuracy and
datasets are presented in section IV and future work and
average recognition time per character. The output system has
been delivered as a website to the end-user with 2 separate conclusion in section V.
portals for customer and parking lot owner. Each portal has its
own features that satisfy its target user. The proposed system
has been efficient in satisfying needs of the customer and the II. LITERATURE SURVEY
vehicle type classification model has given a good accuracy The authors[1] proposes an efficient approach to identify
score. available parking spots and regulating the flow of vehicles in
complex multi-storied parking infrastructures by means of IR
Keywords—ARIMA, LSTM, B-LSTM, Neural
sensors that detect vehicles and provide interpretation. This
Prophet
system can be dejected into metropolitan malls, hybrid
I. INTRODUCTION parking structures.
Automatic car slot identification system[2] works by
Everyone in today's busy world values their time and waiting
displaying vacant parking slot numbers on monitor screens
for anything is not an option. A lot of parking systems present
various difficulties for people such as inefficient tracking of located at the entry gate before drivers even enter the lot. By
vehicles, waste of time caused by arbitrary parking, and implementing this system, drivers can save time and avoid
inadequate parking lot management. This results in the the frustrating and inefficient process of searching for open
significant consumption of time and financial resources, not spaces. Only the credible and legible users can access the
just for the car owners but also for the parking lot owners who parking lots with the usage of RFID tags subject to the
need to compensate their employees responsible for availability of the parking dimensions.
monitoring and managing the parking lot. In addition to that,
it's challenging for parking lot owners to manage their A significant touch of an efficient smart parking management
operations through manual labor, which drives up labor costs. system [3] is proposed and implemented by mainly using
The most effective solution for these issues is implementing techniques such as image processing , hardware and electrical

979-8-3503-0681-1/23/$31.00 ©2023 IEEE

Authorized licensed use limited to: Zhejiang University. Downloaded on November 13,2024 at 03:28:12 UTC from IEEE Xplore. Restrictions apply.
apparatuses. The system avails appealing upgrades, such as to monitor, control, process, and display the required parking
automatic license plate recognition, anti-theft parking space information in real-time to the mobility seekers.
security, and even more innovations.
III. METHODOLOGY
A fully automated parking lot management system[4] using
an embedded robot type smart car that is applicable to any A. EXISTING METHODOLOGY
type of parking lot has been prescribed .The architecture In many parking lots, the conventional method of parking
classifies the vehicles after detecting them into three vehicles involves a manual process. When you enter, the staff
categories: usually parked vehicles, illegally parked vehicles, provides a parking ticket, and you can locate a free space to
park your vehicle without any specific instructions. Staff
and improper parked vehicles. Ultrasonic and infrared
members are present to assist in navigation. When you leave,
sensors are deployed in the smart car. a staff member collects the depicted fee, totted on the basis
A novel model [5] provides an optimal solution only for the of incoming and outgoing times. Although this system has
parking lot issues in cosmopolitan cities. Due to the ongoing some downsides, it remains a commonly used approach in
decent increase in the usage of vehicles in such cities. The parking lots. Even now, many feel this method is comfortable
infrastructure includes a slot module that is located on site as there are not many alternative methods which have been
and is used to monitor and signal the availability of each implemented on a large scale.
An alternate solution to this issue involves the utilization of
parking space. In addition to that, a fully developed
sensors to detect the presence of vehicles at that location. This
application is made available to the users for their tracking
strategy entails the installation of sensors in parking spots,
upon the depicted criteria. This plays a significant role in
which will identify unoccupied spaces within the parking lot.
increasing the economy via the effective infrastructure.
The information gathered by these sensors can be utilized by
The parking system proposed in [6] is based on sensors and
drivers to locate vacant parking spots. While this method has
cloud server integration. In the pre phase, the ultrasonic
been implemented in some parking areas, the primary issue
sensor is used to detect vacant or occupied car parking space
with this strategy is its high cost. The expense of installing
and showcases it on the screen using cloud technology. In the
sensors ranges from $300 to $500 for each parking spot.
post phase, the improper parking of the car is detected by IR
Despite the high cost, some organizations are prepared to
sensors. It warns the transport officer with a buzzer if the car
finance the utilization of sensor-based parking systems.
is not parked in the parking boundary.
The most reliable and efficient way to automate the
management of the parking system [7]that allocates B. PROPOSED SOLUTION
exhaustive parking space by incorporating the internet of A cutting-edge smart vehicle, primarily car parking system
things. was manifested that has preparedness to automatically
In the work [8] ,the proposed ideology utilizes the recent allocate an appropriate parking spot for any vehicle that
advancements in the Information and Communication enters the parking facility. With the aid of a user-friendly
Technologies and consists of four layers: Application, website and a lightweight number plate and vehicle type
Middleware, Networking, and sensor layer. It provides detection system, this Smart Parking System is designed to
environmentally friendly, reduces exhaust emissions during help parking lot owners manage their parking spaces more
parking, and is a mechanized system pre-programmed efficiently. The website also facilitates easy communication
without human influence. between customers and parking lot owners. From assigning a
The IoT provides access virtually to the system and the survival of the fittest parking spot for a vehicle at the
mobility users can keep a track of the availability of the immigrant point to calculating the total parking charge at the
parking area. Road congestion is a time consuming factor in emigrant point, this smart parking system handles it all.
metropolitan cities. The complete information regarding the The automated smart parking system operates without much
parking is sent to the user via an alert. Thus, the waiting time human interaction, and it uses advanced technology to
for the user is minimized. RFID technology is being deployed recognize the number plate and the vehicle type (SUV,
Sedan, or Compact) when a car enters. The system then
to avoid car theft [9,10].
utilizes an effective search algorithm to round about a
To improvise reliable and efficient mobility in cities with an congruous parking extent upon the specific details of the car.
extremely large vehicle fleet, a crucial solution that brings the Depending on the vehicle type, the parking space allotted
cornerstone of a smart parking system into action, the smart may be compact, medium, or large. In addition, if it is the
vehicular presence sensor (SPIN-V, in its Spanish user's first visit, a user ID is generated, and the user is
abbreviation), is presented[11]. The components of SPIN-V provided with login credentials. After entering the parking
include a single-board computer, distance sensor, camera, lot, each parking spot will be monitored individually using a
number plate recognition system. If a customer parks his/her
LED indicator, buzzer, and battery. This smart inno-mobility
vehicle in a spot that is not assigned to him/her, a circumspect
project includes three main elements, namely the SPIN-V, a notification will be sent to the customer and the customer
mobile app[12], and center for monitoring, binding together would be fined if he does not respond to it. At the exit point,
the vehicle's total charge will be calculated based on facts like

Authorized licensed use limited to: Zhejiang University. Downloaded on November 13,2024 at 03:28:12 UTC from IEEE Xplore. Restrictions apply.
entry time and unveiled to the user, so that the authenticated Convolutional Neural Network (CNN) for the detection of
client can proceed with the payment. objects. Fig. 2 infers the architecture of the prescribed YOLO
model.

Fig. 2. YOLOv3 Architecture

Fig. 1. Smart Parking System Workflow
E. VEHICLE TYPE CLASSIFICATION
C. SMART PARKING WEBSITE: ALGORITHM:
The smart parking system comprises two different portals, A subset of Stanford Car Body Classification dataset has been
namely the admin portal and the customer portal, both of used for classification of the vehicle type. The Stanford Cars
which have unique features designed to assist their intended text file concludes 196 car classes summing up to a total of
users. The admin portal, for instance, provides a range of 16,185 images, propelled from the rear. The train subset of
functionalities such as creating and managing new parking the Stanford dataset has been used for the creation of our
lots. Additionally, admins can establish multiple parking lots dataset. It consists of images of cars categorized according to
within the same organization or across different their body type like SUV, Sedan, Hatchback, Van, Cab,
organizations, with each parking lot being mapped to a Coupe, Convertible, Minivan, etc., From these car types,
specific organization. The workflow of smart parking system images of SUV and Sedan has been extracted and used to
is depicted in Fig 1. solve our vehicle type classification case. The extracted
dataset has about 2017 images of Sedan and 1767 images of
D. VEHICLE DETECTION AND CLASSIFICATION SUV.
At the entry point, the vehicle’s type would be classified to F. VEHICLE NUMBER PLATE RECOGNITION
assign a suitable spot for the vehicle. The vehicle type Vehicle Number Plate Recognition system is used both at
classification has been done with the help of two models (i.e., entrance and exit as the vehicle number has been used as a
a vehicle detection model and a vehicle classification model). unique id to track the vehicle. It is also used to monitor the
A camera would be fixed at the entry and exit point to capture vehicle’s activities inside the parking lot which could be
live video feed of incoming vehicles. The cameras would be helpful in alerting the respective user if there is any need. The
connected with a raspberry pi through IOT, and the light- vehicle number is used to calculate the parking charge at the
weight vehicle detection and classification models would be exit point based on the immigrant and emigrant time of the
loaded in the raspberry pi. Frames from the camera would be automobile. The process carried out for vehicle number plate
extracted at a rate of 30 fps and passed on to the vehicle recognition has been explained in the upcoming part.
detection algorithm which detects the vehicles present in the Vehicle Number Plate Detection can be performed using
frame. The detected parts would then be cropped out and Cascade Classifier. The detection of vehicle’s number plate
passed on to the next model which classifies the vehicle type. is done using a Haar-Feature based Cascade classifier model.
The vehicle classification model would output the vehicle’s Cascade classifier is a prior trained classification which is
type which would be passed on to the search algorithm to find trained using several positive images and arbitrary negative
a suitable parking spot. This outputs the corresponding images. The pretrained classifiers are stored as separate xml
vehicle’s type as output which is passed on to the search files in OpenCV package. In this case, xml file of the Indian
algorithm to find a suitable parking spot that can number plates has been used for detection of number plate.
accommodate the vehicle. The extracted car’s image is passed on to the cascade
Coco dataset has been used to train the Yolov3 model which classifier which in turn detects the number plate and further
is used for vehicle detection. The MS COCO dataset is for processing is done to extract and classify the characters in the
distinguishing wide-scale objects, segmentation, key-point number plate.
detection, and captioning the accumulated dataset. The G. Character Extraction from License Plate:
dataset contains about 328,000 images of the objects that we
come across in our day-to-day life, each dotnotted with The license plate detection involves some preliminary pre-
categories of concluding 80 objects and 5 captions in processing steps to extract characters from the received
numbers. image. These steps include converting the image into
grayscale format followed by threshold application to
YOLO model has been trained for vehicle detection at entry highlight the characters and differentiate them from the
and exit points. YOLO is a light-weight object detection background. This helps the algorithm to detect the character
model that is used for analysis of objects in a realistic video borders accurately. The image is then subjected to erosion
or in an image. The prediction that is used in YOLO is based and dilation to remove noise from it. After contour detection,
on a convolutional layer that uses 1×1 convolutions. It uses a contours are extracted from the image and the top 15 highest
contours are sorted to segment characters. The segmented

Authorized licensed use limited to: Zhejiang University. Downloaded on November 13,2024 at 03:28:12 UTC from IEEE Xplore. Restrictions apply.
characters are finally passed on to the optimal character B. VEHICLE TYPE CLASSIFICATION:
recognition model. This character recognition model would Vehicle Type Classification process has been tested with 3
classify/recognize the characters. The individually models (EfficientNet,based TL Model, MobileNet, light
recognized characters would be joined to give the vehicle weight CNN model). Table. 2 shows the accuracy, loss and
number as output. average classification time of different models tested for
vehicle type classification. The accuracy score of the efficient
IV. EXPERIMENTAL RESULTS net based model is about 96.6% whereas the mobile net
A. VEHICLE DETECTION: extended model, light weight CNN model,DenseNet-201
extended model and InceptionV3 based model gives an
Vehicle Detection has been tested using Yolov3 and Tiny
accuracy of about 91.9%, 72%,92.2% and 75.8%
Yolov3 models. The test results of the models has been
respectively. Efficient based TL model has loss value of
shown in Table 1. In summary, both the models have done
0.1516 which is less compared to Mobile Net Extended
reasonably well in real-time videos. However, it is evident
model and light-weight CNN model which have loss values
that tiny yolo was much faster in vehicle detection than yolo
of about 0.1892 and 0.789 respectively. The average
which is an added advantage for the model. We could see that
classification time of EfficientNet based TL model is about
the tiny yolo model takes an average detection time of 0.121
0.269 seconds while the MobileNet extended model, CNN
seconds per frame while yolo takes about 0.656 seconds per
model,DenseNet-201 extended model and InceptionV3
frame. As the model is to be implemented for real-time
model takes about 0.274 seconds, 0.279,0.324 and 0.5379
predictions, the model is required to be much faster and so in
seconds respectively.
that way tiny-yolo is much better. On the other hand, yolo
Two of the best adaptive optimizers (Adam and RMSProp)
model gives an accuracy of about 67% which is better than
has been tested using along with Efficient-based transfer
tiny yolo model which gives an accuracy of 54%.
learning model to decide on the best one. The model gave
Performance Analysis of Vehicle detection algorithms is
particularly good accuracy of about 96.1% when tested with
given in Table 1 and results in Fig 3.
Adam Optimizer and gave an accuracy of about 94.3% with
Table 1. Performance Analysis of Vehicle detection
RMSProp Optimizer. As the Adam Optimizer gave better
algorithms
accuracy and performed well in real-time traffic video, it has
Model Accuracy(In F1 Score(In Detection
been chosen for the model.
real-time real-time Time (per Table 2. Accuracy and Loss of EfficientNet-based based
Data) data) frame) model with different optimizers
Yolov3 67% 68% 0.656 secs
Tiny 54% 56% 0.121 secs Model Validation Validation
Yolov3 loss accuracy
Adam Optimizer 0.1516 96.6%
RMSProp Optimizer 0.1675 94.8%

The vehicle type classification has been tested with

real-time videos and the model performed well in it. Fig. 4
shows the output screenshots of the model tested with real-
time video.
(a)

(a) (b)
(b)

(c )
(c)
Fig. 3 -a,b,c Results of Model Tested in real-time video

Authorized licensed use limited to: Zhejiang University. Downloaded on November 13,2024 at 03:28:12 UTC from IEEE Xplore. Restrictions apply.
Fig. 4 Vehicle Type Classification Output In Real-time video on the complexity of the model.Table 3 and 4 illustrates the
Model Accurac Loss Average performance of various models.
y Classificati
on Table 4 . Accuracy and Loss of models with respect to test dataset
Time(per
frame) Model Accuracy Loss Classification
EfficientNet 96.6% 0.1516 0.269 s Time(per
character) in secs
based transfer
learning CNN Model 95.6% 0.149 1.143
model ResNet50 96.3% 0.129 1.332
MobileNet 91.9% 0.1892 0.274 s Extended
based transfer Model
learning ResNet152 96.9% 0.108 3.253
model Extended
CNN based 72% 0.54 0.279 s Model
model VGG16 91.7% 0.285 2.502
DenseNet 201 92.2% 0.2529 0.324 s Extended
based Model
Transfer VGG19 91.9% 0.221 2.413
Learning Extended
Model Model
InceptionV3 75.8% 0.5379 0.271 s
based transfer The CNN Model used for Number Plate Recognition has
performed well and it has been noticeably light weight which
learning
would be helpful for a real-time application. The model gave
model an accuracy of about 95.6 % in test dataset which had
C. VEHICLE NUMBER PLATE RECOGNITION: different strokes of the number. The test loss was about 0.149.
The cascade classifier model used for number plate It took about 700 seconds to train with the dataset which is
recognition has been good at recognizing the number plates reasonably less in terms of training time for the used epochs.
Table 3. Accuracy and Loss of models with respect to test dataset Hence the model is less complex and performed extremely
in real-time images. The cascade classifier model is much well in real-time images. Fig.6. depicts the out-turn predicted
faster, and it has been very useful while used with live stream of the optimal character recognition model.
videos in parking lot system. Fig. 5 shows the number plate
detected by cascade classifier model in a real-time image .
The cascade classifier model is accurate enough in detecting
number plate that is present in a vehicle present at a good
distance.

Fig 6 , Character Extraction

Fig. 5 Number Plate Detection Using Cascade Classifier Model In summary, The EfficientNet based Transfer learning model
The data has been tested with different models and which has been trained on the Stanford Car Body Dataset for
a model that is faster and more accurate in prediction has been the purpose of vehicle type classification at entry point has
chosen. Table 3 prescribes the metrics including accuracy, performed well in test dataset as well as real-time images in
precision and recall scores of different models trained on the comparison to other models. On the other hand, among the
characters dataset. In summary, we could see that the models performed trained for number plate recognition with
accuracy scores of CNN, ResNet50, ResNet152,VGG16 and standard OCR dataset, the CNN model was faster in
VGG19 models are 95.6%,96.3%,96.9%,91.7%,91.9% recognizing the character and the ResNet152 was more
respectively. In terms of classification time, the ResNet152 precise in prediction. As the model's accuracies does not vary
model took about 3.253 seconds to classify a character which much, the decision was taken based on the recognition time
is comparatively higher than that of ResNet50, and complexity of the model. The vehicle detection models
VGG16,VGG19 and CNN Models that took about 1.332, that have been trained on COCO dataset has been compared
2.502,2.413 and 1.143 seconds respectively. As the CNN by testing the accuracies and detection time in real-time
model is light weight and faster in classification, it video. YOLO model has given a better accuracy score in real-
outperforms other models and as the accuracies of the models time data and the tiny YOLO model has been faster in
does not vary much, the decision has been finally taken based detection. The models support in automation of parking

Authorized licensed use limited to: Zhejiang University. Downloaded on November 13,2024 at 03:28:12 UTC from IEEE Xplore. Restrictions apply.
system for unique identification of a vehicle and Information Processing (ISSNIP), 2015 IEEE Tenth
identification of the vehicle type to decide on a suitable spot. International Conference on (pp. 1-6). IEEE.
[10] Rico, J., Sancho, J., Cendon, B., & Camus, M. (2013,
V. CONCLUSION: March). Parking easier by using context information of a
A more effective fully automated smart parking system that smart city: Enabling fast search and management of parking
delivers a client-friendly website to the end user, along with resources. In Advanced Information Networking and
some automations at the entry and exit points has been Applications Workshops (WAINA), 2013 27th International
proposed. The system uses yolov3 model to detect vehicles Conference on (pp. 1380-1385). IEEE.
in real-time which is then passed on to the transfer-learning [11] N. Hazrin, H. Mohamad, M. H. Badiozaman, and H.
model (with EfficientNet-B2 as base layer) to classify the Daud, “Smart Parking Reservation System using Short
vehicle type. The EfficientNet based transfer-learning model Message Services (SMS),” 2008.
which has been proposed for vehicle type classification [12] R. Renuka and S. Dhanalakshmi, “Android Based Smart
leverages accuracy of 96.2% is higher when compared to the Parking System Using Slot Allocation & Reservations,” vol.
other models. Even though the proposed model is light- 10, no. 7, pp. 3116–3120, 2015.
weight and works well in real-time, there is still a scope for [13] T. N. A. M. Pham, M. Tsai, and D. U. C. B.
improvement. The system uses cascade classifier model to Nguyen, “A Cloud-Based Smart-Parking System Based on
recognize the number plate and it uses a CNN based OCR Internet-of-Things Technologies,” pp. 1581–1591, 2015.
model to identify the vehicle number. The OCR model gives [14] M. Computing, “Smart Car Parking Using Arduino,”
an accuracy of 95.6% in character recognition, and it vol. 5, no. 2, pp. 230–234, 2016.
performs well in real-time. In the future, it could also be used [15] Prof. D. J. Bonde , Rohit S. Shende, Ketan S. Gaikwad,
to manage multiple parking lots from various organizations Akshay S. Kusarigama U. Bhakra. “Automated Car
in a single site, enabling prospects to probe the existence of Parking System Commanded by Android Application”,
parking spaces in any parking lot using a single app or site. (IJCSIT) International Journal of Computer Science and
Information Technologies, volume 5(3), pp. 1-4, 2014.
[16] Dhaou, I. B., Kondoro, A., Alsabhawi, A. H.,
REFERENCES: Guedhami, O., & Tenhunen, H. A Smart Parking
[1] Thanh Nam Pham1, Ming-Fong Tsai1, Duc Binh Management System Using IoT Technology. In Proceedings
Nguyen1, Chyi-Ren Dow1, And Der-Jiunn Deng2 “A Cloud- of the 22st Conference of Open Innovations Association
Based Smart-Parking System Based on Internet-of-Things FRUCT, 43 (2018).
Technologies”,IEEE Access, Received July 24, 2015, [17] S.Poornimakkani, Sushmitha Senthilkumar, S.Finney
accepted August 16, 2015, date of publication September Daniel.. A cloud based end-to-end smart parking solution
9, 2015, date of current version September 23, 2015. powered by IoT. International Research Journal of
[2] Mr. Basavaraju S R “Automatic Smart Parking System Engineering and Technology, 3(5), 3559-3565 (2018).
using Internet of Things (IOT)”, (International Journal of [18] James Gerdeman, “RFID Changing Gates”, IEEE
Scientific and Research Publications, Volume 5, Issue 12, Potentials, vol. 34, pp. 40-42, September 2015.
December 2015) [19] Zhao Du, Yeming Tang, “Web-based multi-level
[3] Hilal Al-Kharusi, Ibrahim Al-Bahadly, “Intelligent smart card access control system on university campus”,
Parking Management System Based on Image Processing”, IEEE International Conference Software Engineering and
World Journal of Engineering and Technology, 2014, 2, 55- Service Science, pp.1015-1018, October 2014.
67. [20] Anusha, Arshitha M, S, Anushri, Geetanjali
[4] Khanna A, Anand R. “IoT based smart parking system”, Bishtannavar “Review Paper on Smart Parking System,”
International Conference on Internet of Things and International Journal of Engineering Research & Technology
Applications (IOTA) (2016 Jan 22): pp. 266-270, IEEE. (IJERT), ISSN: 2278-0181, Volume 7, Issue 08, Special
[5] Subashini, T. S., and Mr. K. Subramanian. "Ticket- Issue – 2019.
based Smart Parking System using Mobile Application in
Vehicular Network." European Journal of Molecular &
Clinical Medicine 7.3 (2021): 685-695.
[6] Yan, Gongjun, et al. "Smart Parking: A secure and
intelligent parking system." IEEE intelligent transportation
systems magazine 3.1 (2011): 18-30.
[7] Juhi Seth, Pola Ashritha, R Namith, “Smart Parking
System using IoT ElakyaR”, International Journal of
Engineering and Advanced Technology (IJEAT), ISSN: 2249
– 8958, Volume-9 Issue-1, October 2019.
[8] Yashomati R. Dhumal, Harshala A. Waghmare,
Aishwarya S. Tole, Swati R. Shilimkar,” Android Based
Smart Car Parking System” Proc., of. IJREEIE, Vol. 5, Issue
3, pp-1371-74, mar-2016.
[9] Zheng, Y., Rajasegarar, S., & Leckie, C. (2015, April).
Parking availability prediction for sensor-enabled car parks
in smart cities. In Intelligent Sensors, Sensor Networks and

Authorized licensed use limited to: Zhejiang University. Downloaded on November 13,2024 at 03:28:12 UTC from IEEE Xplore. Restrictions apply.