Parking Space Monitoring – A LoRa Based

2023 11th International Conference on Internet of Everything, Microwave Engineering, Communication and Networks (IEMECON) | 978-1-6654-7512-9/23/$31.00 ©2023 IEEE | DOI: 10.1109/IEMECON56962.2023.10092339

Approach
Mallikarjun B.C. Anantha Krishna G Mahantesha B.V.
Dept. of ETE Dept. of ETE Dept. of ETE
Siddaganga Institute of Technology Siddaganga Institute of Technology Siddaganga Institute of Technology
Tumkur, Karnataka state, India Tumkur, Karnataka state, India Tumkur, Karnataka state, India
mallikarjun_bc@sit.ac.in 1si18te003@sit.ac.in 1si18te022@sit.ac.in

Monish K.T. Kishore E

Dept. of ETE Dept. of ETE
Siddaganga Institute of Technology Siddaganga Institute of Technology
Tumkur, Karnataka state, India Tumkur, Karnataka state, India
1si18te025@sit.ac.in 1si18te020@sit.ac.in

Abstract— A huge growth of big cities, the number of be the value set else it will be set to 0. In computer vision, this
vehicles are increased significantly and large buildings and method of thresholding is executed on grayscale images. So to
parking lots have been built. Smarter parking systems need to begin with, the photo has to be converted in grayscale shade
be developed to assist drivers in finding nearby parking spaces space [3].
available. Various research efforts have been shown in
developed countries in such system implementation. There are The authors in [4], discuss the technique used for
four main types of systems of parking management: wired improving the detection of vehicle is by making use of
sensor based, image based, counter based, and based on wireless background subtraction algorithms. This procedure
sensors. This paper combines computer vision and LoRa eliminates unwanted parts and more accurately distinguishes
wireless technology and worked on image-based system to find the vehicle. Next, it tracks the vehicle in every frame
empty slot in the parking lot. Following the first edge detection depending upon the information fetched from the earlier
category, includes front/rear pixel rate, closed contour density,
and edge density. For all car parks to determine if the vehicle
frame. The results of experimentation are obtained in
exists or not and to update this information to the user using OpenCV, shows that the accuracy rate reaches nearly 97%
wireless technology in a way that saves power and performance for detecting objects and 98% for tracking objects.
offline while minimizing system maintenance costs.
It is also, vehicles travelling towards the same point in
Keywords—Vacant Parking Space Detection, Image either light or dark areas, or in other way around, may have
Processing, LoRa , OpenCV, Node to Node the same color because the background remains more
difficult to detect the vehicle. Yes, this is observed to lead to
I. INTRODUCTION
vehicle lead count errors [5]. The solutions to this was to use
Nowadays vehicles usage are increasing rapidly so much a background subtraction technique to register as a vehicle
so that search for an empty parking space is a challenge when the specified threshold was exceeded.
especially in big parking lots such as shopping, airports and
railway stations areas. some parking regions may be heavily A Parking slot will be considered as occupied when the
used while others may have more vacant slots. rectangle of that slot has high threshold pixels almost
Due to the heavy traffic and time required to find a parking throughout the outline. This indicates that there is a vehicle in
space, there is a need of a parking management system. that particular space, rectangle of that parking slot will be
Regularly there are two methods used for empty space turned to red colour. Each and every frame of video will
detection: image-based sensor supported systems. Image undergo grayscale and thresholding such that the parking lot
based systems have proven are cost effective and highly is been monitored constantly [6].
durable than sensor based parking systems[1]. The photo Mochamad Mobed Bachitar[7] developed parking
based parking space detection system can notify the driver
information system which delivers the information of free
directly through a separate display panel on the door of each
parking lot location using modern navigation system and also
parking slot or via a mobile phone [2].
used Haar cascade classifiers which is effective for object
Parking detection process tells how many free parking detection based on machine learning. The various input
slots are available or how many are occupied. The parking image cascading features will be in the form of both positive
detection operation is achieved by the image processing and negative images. The positive images are the target
concept called “thresholding”. Thresholding is a totally objects which are need to be detected. Negative images are
famous segmentation approach, used to distinct an object other objects which are not to be considered for detection.
taken into consideration as a foreground from its background. The HAAR Wavelet is compared to the average of two pixels.
A threshold is a number which has two domain on its either The compared value of average intensity is greater than the
aspect i.e. above or below the threshold. In thresholding, the threshold, then it is concluded that the target object is
threshold value is compared with each pixel value. If value of
occupied.
the pixel is greater than the required threshold, then 255 will

978-1-6654-7512-9/23/$31.00 ©2023 IEEE

Authorized licensed use limited to: SIDDAGANGA INSTITUTE OF TECHNOLOGY. Downloaded on July 16,2024 at 04:34:05 UTC from IEEE Xplore. Restrictions apply.
 The authors in article [8], describes the parking devices are sending wireless data and receiving
surveillance system which is designed to detect the acknowledgments at the same time.
availability of free space during lightened and complicated
situations. System make use of fixed camera for monitoring The primary aim of the system is to detect free parking
the small parking area by using captured raw original images. space, where it retains the count of available parking slots in
The proposed system used some major steps they are (1) the certain monitored area. Raspberry pi 3 and camera module
Spatial Initialisation : the system is initialised by capturing of raspberry pi is used to monitor vehicles, where OpenCV
targeted parking space's image and this image should python program is dumped.
correspond to an image which is taken when parking slots are
empty, (2) Manually select the scalable rectangles over Compared to other vehicle detection technology, this
reference image, (3) Then by using image based measuring proposed system has easy installation and maintenance, cost
mechanism slot occupancy is calculated using error function effective and uses LoRa for transmission of monitored data to
and (4) Consequently self-adjusting mechanism allows to the user destination without internet.
refine occupancy threshold even when there is imbalance in
II. PROPOSED SYSTEM
occupied slot and vacant slot .Therefore the vacant spaces are
coloured in green and occupied space in red according to their The proposed system is very compact and more power
threshold value. efficient than the traditional sensor-based system, In this
system, we completely eliminated the wired communication
LoRa is low power wide area wireless technology used to
by using LoRa wireless communication which has a very
for machine to machine communication with long range
capability, low power for data transmission and offers secure long range and consumes less power to operate and gives an
data transmission. LoRa uses chirp spread spectrum(CSS) edge over the traditional sensor-based system. Each parking
modulation technique where it utilizes the whole bandwidth area consists of a processing node with a camera, Raspberry
of the channel to transmit a signal which offers resistance to pi, and a Lora transmitter node, which detects the free slot in
noise in the channel, fading and doppler effects, making it a parking area and sends results through the Lora module to
feasible to reach very large communication distances[9]. a centralized location, data is monitored and broadcast to the
Compared to other data transmission protocols, LoRa requires user using IoT.
very less power, conserving the battery life of devices which
can run for several years on the same battery. The frequency A Lora Gateway is located in the central part of the
at which the device operates don’t require licensing such that surveillance area which is the receiver part of the system
deployment costs are economically viable [10]. equipped with an internet connection and receives the
In [11], LoRa devices in open area can send data upto 15 information from Lora nodes using Lora communication. The
km, this is enough for any long-distance IoT applications data from many nodes are analyzed and broadcasted to the
today. However, LoRa devices may not be able to end user, a typical Gateway can handle thousands of these
communicate wirelessly in indoor applications with nearby nodes over a 1Km-5Km radius depending upon the data rates
Gateway, due to objects in between the sensors, this reduces which reduces the maintenance cost over the sensor-based
the strength of the radio signal and can lead to data loss and system. A centralized system is more efficient than a
communication errors. decentralized system thereby reducing the processing power
at each level
In [12], the Spreading Factor (SF) finds out the number of III. SYSTEM IMPLEMENTATION
data carrier symbols are sent per second. The network
determines the diffusivity (range 7 to 12) depending on the Before implementing the proposed system, initial
installation & standardization is required. For the provided
environmental conditions between the communication device
image covering a specific parking area, camera position
and the gateway. This requires the adaptive data rate feature
should be adjusted. The boundaries of the parking space need
to be turned on. This is recommended for all devices except to be fixed. Then, we can define the location of each parking
those that are always on the move. A low SF means that more lot using 4 dots or parallelogram for a given parking space.
chirps are sent per second. Therefore, you can encode more Here, each parking space is assigned a number and usually
data per second. The higher the SF, the less chirp per second. numbered from the top number. Once the installation and
Therefore, less data needs to be encoded per second. Sending measurement phases have been performed, the system is
the same amount of data at a higher SF compared to a lower active and ready to use.
SF results in a longer transmission time, called airtime. More
airtime means that the modem will run longer and use more
power.

In [13] experiment result shows that increasing the

spreading factor (SF) to 12 can greatly expand the
communication distance by maximising the sensitivity of the
receiver . However, the disadvantage is discussed in [14] that
it slows down the data throughput rate and increases the
transmission time, which can lead to more serious data Fig 1. System overview
collisions. In [15] experiment results shows that this issue can
be exacerbated when a large number of high-density LoRa Since video is captured from raspberry pi camera module
we need to check the camera resolution. The height and width
should be entered in the code as per the camera module

Authorized licensed use limited to: SIDDAGANGA INSTITUTE OF TECHNOLOGY. Downloaded on July 16,2024 at 04:34:05 UTC from IEEE Xplore. Restrictions apply.
 resolution. The system will fail to produce the output if there A. Detection by Edge Pixel Counting
is discrepancy. The resolution setting therefore varies with the We begin with edge detection algorithm using the
adjustment of each camera. 640 x 480 is the recommended Prewitt edge operator, shown in Table 1:
resolution for raspberry pi camera module to get optimal
performance. even higher resolution like 1080p and 720p are Z1 Z2 Z3 -1 -1 -1 -1 0 1
supported. Z4 Z5 Z6 0 0 0 -1 0 1
Z7 Z8 Z9 1 1 1 -1 0 1
Image Neighborhood Mask of Gx Mask of Gy

Table 1 Prewitt operator

The edges are calculated using the difference between the

corresponding pixel intensity of the image. All masks used
for edge detection are also known as mask extensions.
Because image is also a signal so changes in the signal can
only be calculated using divisions. This is why these
operators are also referred to as outgoing operators or
outgoing masks.

Approximation of first order derivative is obtained using

Prewitt operator. For a particular pixel at gray level G(i, j),
Overall, the strongest values are the maximum values for Gx
and Gy, and if this value is greater than certain limit, it is
declared as an edge pixel.

Once the Prewitt edge detector is applied, parking space

filter is used to extract the edge information of every parking
slot and then count the number of edges. Space filter only
encloses the edge pixels inside the predetermined car slots
obtained from the monitoring area. Suppose the total number
of edge pixels is smaller than definite value, the parking slot
is considered as free.
B. Detection by Foreground/Background Information
In this the image goes through following processes:
• Convert the input frame to Gray scale image.
• Histogram is obtained from Gray scale image.
• The bin with maximum frequency is considered as the
background bin and those pixel are taken as
background pixels. while extracting the foreground
data, mean of all pixels relative to background is taken.
• The difference between the grey level frame and the
background frame is determined. Non-zero pixel
values gives data on whether a particular slot has
Fig 2. Flow chart of Methodology occupied by a car or just a background.
Before deploying or installing the system certain • Number of non-zero pixels from previous step of each
calibration should be done. There are many positions in a parking space is calculated and compared to the
given site for the raspberry pi camera to be fixed. Each camera threshold, if the number exceeds the limit, the space is
position has different view towards the monitoring area and called as busy, otherwise as free.
when the system is used for different locations. the parking
layout varies from place to place. So to have effective parking
detection in all location and scenario. We will be manually
outlining every parking slot for a given area by drawing
rectangle on a image representing the surveillance view of the
area while installing the raspberry pi module. by following
this method all the parking slots will be monitored effectively
and efficiently provided any camera position and place.
Outline colour of parking slots will turn green if it is
unoccupied and will turn red if it is occupied by a vehicle. The
camera should be above certain height such that entire parking Fig 3.flow diagram of data transmission
area is covered in the camera frame. Whenever there is state change in parking area the results
is transmitted to the user using LoRa wireless communication.

Authorized licensed use limited to: SIDDAGANGA INSTITUTE OF TECHNOLOGY. Downloaded on July 16,2024 at 04:34:05 UTC from IEEE Xplore. Restrictions apply.
 The LoRa module will serial communicate with the raspberry We extract the edge information for each parking slot the
pi and sends the data to receiver node without internet over a result of the edge detection is shown in Table 2.
long distance through unlicensed radio frequency (868 MHz)
whose range is nearly 5kms in urban and 15kms in suburban Table 2 results of Edge detection technique
areas. Space1 Space2 Space3 Space4 Space5
The receiver node receives the information and using Edge 381 280 559 703 315
internet the results are updated to could through which the user pixel
Decision occupied free occupied occupied free
can view the data in mobile application.
From many iteration we carried on fully busy parking area,
the result showed us the threshold of 336 for the edge
IV. RESULTS
counting technique.
From experimental analysis, the proposed system was
tested on a bypath background. The recording were taken For identifying the background pixels, we use histogram
with bird view type of camera adjustment covering 5 to 7 as shown in figure 7.
parking spaces. An example of the parking view is shown in
Figure 4.

Fig 7. Histogram of input frame

Fig 4. Input frame The maximum frequency of histogram is seen at pixel
168-179. Pixel values lies in between this interval are taken
During implementation and calibration, each and every as background pixels. We consider mean values of these
parking slots are segmented and numbered as shown in pixels corresponding to background values. Then the
Figure 5. foreground pixels is extracted using this background
information. The foreground Mean Pixel Value (FMPV) are
displayed in table 3

Table 3 results using foreground/background method

Space1 Space2 Space3 Space4 Space5
FMPV 132 39 116 108 36
Decision occupied free occupied occupied free

When we carried out many iteration of vacant parking

area, the results showed that the threshold of 47 is suitable to
identify free and occupied slot. The result shown on table 3
and 4 are an image covering 5 parking space captured in
Fig 5. Segmented output raspberry pi camera. Note that the threshold value varies with
After the edge detection processes, the result images is different camera used.
shown in Figure 6
The information from the raspberry-pi is send to user by
using Lora modules the transmitter node will serial
communicate with raspberry-pi and collects the information
and transmit the data to receiver node without internet over a
long distance through unlicensed radio frequency (868 MHz).

The range between two nodes can be increased by make

making the spreading factor to 12. This results in low data
rate but long range as our information is less in size this is
suitable for us .The receiver node has internet connection
through which we can able to upload the results to could and
Fig 6. Edge detection of input frame user can actively monitor the parking area.

Authorized licensed use limited to: SIDDAGANGA INSTITUTE OF TECHNOLOGY. Downloaded on July 16,2024 at 04:34:05 UTC from IEEE Xplore. Restrictions apply.
 V. CONCLUSIONS [6] Parking Space Detection in OpenCV: May 29, 2018 [Online].
Available: Parking Space Detection in OpenCV | ParkingLot
We developed and tested a economical vigorous vision (olgarose.github.io), 2018
based parking space monitoring system. The parking state is [7] Mochamad Mobed Bachtiar,Adnan Rachmat Anom Besari,Atikah
determined in real-time using fast and robust image-based Putri Lestari-Parking Management by Means of Computer Vision.The
third International Conference on Vocational Education and Electrical
classifiers. Which could be installed in different parking Engineering (ICVEE) 2020.
layouts. The results attained from testing the image (video [8] Karim Hammoudi,Halim Benhabiles,Abhishek Jandial-Self-Driven
frames) based parking system, indicates that the likelihood of and Direct Spatio-temporal Mechanisms for the Vision-based Parking
effective utilization of a vision based or image based system Slot Surveillance.SAI Computing Conference ,July 2016.
to monitor vehicles is feasible. [9] Shilpa Devalal and A Karthikeyan, “LoRa Technology -An Overview”.
However, there are some drawbacks of this system. When Second International Conference Electronics, communication and
Aerospace Technology (ICECA), 29-31, Coimbatore, India, March,
the color of the background and vehicle are similar, detection 2018.
of foreground/background becomes difficult. Usage of [10] Umber Noreen, Ahcène Bounceur and Laurent Clavier “A study of
raspberry pi camera module shows some limitations in terms LoRa low power and wide area network technology” International
of precision and detection. In order to overcome this Conference on Advanced Technologies for Signal and Image
limitation observed in development phase, raspberry pi Processing (ATSIP) , 22-24 May 2017.
camera module can be replaced with much upgraded cameras [11] LoRa Modulation Basics. Accessed: Sep. 3, 2017. [Online].
Available:http://www.semtech.com/images/datasheet/an1200.22.pdf,
like CCTV camera or other superior cameras. By using a Lora 2017.
gateway many parking area in around area of 5km radius can [12] Spreading Factor affects LoRaWAN® device battery life [online]
be monitored and use this information a smart city can available: https://www.thethingsnetwork.org/article/how-spreading-
manage the parking problem well. factor-affects-lorawan-device-battery-life
[13] M. Centenaro, L. Vangelista, A. Zanella, and M. Zorzi, “Long-range
REFERENCES communications in unlicensed bands: The rising stars in the IoT and
smart city scenarios,” IEEE Wireless Commun., vol. 23, no. 5, Oct.
2016.
[1] S. Kaplan, S. Bekhor, S. Kaplan, and S. Bekhor, “Exploring En-route
Parking Type and Parking-search Route Choice‖,” Proc. 2nd Int. [14] Huang-Chen Lee,Kai-Hsiang Ke-Monitoring of Large-Area IoT
Choice Model. Conf., 2011. Sensors Using a LoRa Wireless Mesh Network System: Design and
Evaluation.IEEE TRANSACTIONS ON INSTRUMENTATION
[2] K. Blumer, H. R. Halaseh, M. U. Ahsan, H. Dong, and N. Mavridis, AND MEASUREMENT, 2018.
“Cost-Effective Single-camera Multi-car Parking Monitoring and
Vacancy Detection Towards Real-world Parking Statistics and Real- [15] Smangaliso Mnguni, Pragasen Mudali, Adnan M. Abu-Mahfouz &
time Reporting‖,” in Proceedings of the 19th international conference Matthew Adigun “Performance Evaluation of Spreading Factors in
on Neural Information Processing - Volume Part V, Berlin, Heidelberg, LoRa Networks”, December 2-4, 2020.
pp 506–515, 2012.
[3] Junzho Liu, Mohamed Mohandes, Mohamed Deriche – Multi-
Classifier Image Based Vacant Parking Detection System, IEEE 2013
[4] K apeksha P Kulkarni, Vishwanath P Baligar,- Real Time Vehicle
Detection, Tracking and Counting Using Raspberry-pi,Proceedings of
the Secound Internatinal Conference on Innovative Mechanisms for
Industry Applications, ICIMIA, 2020.
[5] Robert,K (2009). Video-based traffic monitoring at day andnight
vehicle features detection tracking. 12th International IEEE Conference
on Intelligent Transportation system, 2009.

Authorized licensed use limited to: SIDDAGANGA INSTITUTE OF TECHNOLOGY. Downloaded on July 16,2024 at 04:34:05 UTC from IEEE Xplore. Restrictions apply.