Embedded System Design of a Real-time Parking

Guidance System
Omkar Dokur Srinivas Katkoori and Nouredddine Elmehraz
Department of Electrical Engineering Department of Computer Science and Engineering
University of South Florida, Tampa, FL 33620 University of South Florida, Tampa, FL 33620
Email: [email protected] Email: {katkoori, nelmehra}@mail.usf.edu

Abstract—The primary objective of this work is to design (SGEF). This paper focuses on designing of an embedded
a parking guidance system to reliably detect entering/exiting system that detects vehicles when they enter/exit the parking
vehicles to a parking garage in a cost-efficient manner. Existing garage. This system is part of an overall parking guidance
solutions (inductive loops, RFID based systems, and video image
processors) at shopping malls, universities, airports etc., are system that can facilitate the drivers to know the available
expensive due to high installation and maintenance costs. There number of spaces in a parking lot/garage before entering via
is a need for a parking guidance system that is reliable, accurate, a smart phone app. The proposed system connects to the web
and cost-effective. The proposed parking guidance system is server over wired internet connection and updates the server
designed to optimize the use of parking spaces and to reduce wait with the parking space count in real time. The companion
times. Based on a literature review we identify that the ultrasonic
sensor is suitable to detect an entering/exiting vehicle. Initial smart phone app connects to the web server and obtains the
experiments were performed to test the sensor using an Arduino parking count information via web query (HTTP REST API
based embedded system. Detection logic was then developed to and JSON format). The queried data is presented to the user
identify a car after analyzing the initial test results. This logic was in the smart phone app.
extended to trigger a camera to take an image of the vehicle for
validation purposes. This system consists of Arduino, ultrasonic A. Motivation
sensor, and a temperature sensor. It was installed and tested in
Richard Beard Garage at the University of South Florida for five The University of South Florida (USF) has four parking
days. The test results of each trial are provided and average error garages and eighty eight surface lots providing 20,000 parking
for all the trials is calculated. The error cases occur due to golf spaces. However, students and faculty still find difficulty in
carts, straddling cars on both entry/exit lanes, and people walking parking their vehicles during rush hours leading to excess
under the sensor. The average error of the system is 5.36% over
five days (120 hrs). The estimated cost for one detector per lane fuel consumption. The Center for Urban Transportation and
is approximately $30. Research (CUTR) surveyed (unpublished) three parking lots
Index Terms—Arduino, Ultrasonic sensor, Vehicle detection, (CBG, 29A, 29B) at USF to estimate the average time taken
Space availability by drivers to park their vehicle. On average, each student
takes 3.76 minutes to park their vehicle in the above three
I. I NTRODUCTION garages which produces approximately 332 metric tons of
Parking a car is becoming a challenge to drivers. Drivers CO2 emission annually into the campus environment. Fuel
are frustrated while waiting for a long time in a queue or will and time can be saved by informing users about the space
just drive through the entire parking lot but still unable to availability well in advance. With a reliable parking guidance
find a parking spot [1]. While searching for a vacant parking system, we can reduce the time taken by students to park
spot, fuel is lost, and time is wasted; this also increases the and decrease CO2 emissions, thus protecting the environment.
traffic due to slow moving vehicles that are driving around According to the CUTR surveys, if the parking time is reduced
the parking garage [1]. Further it contributes to green house by 2 minutes, the CO2 emissions would be reduced from
gas emissions (CO2 ). A study from Boston University states the three garages to 155 metric tons (53% annual reduction).
that more than 30% of the drivers take around 7.8 minutes to Figure 1 is the preliminary estimation done by CUTR. This
park their vehicle [2]. Parking spaces are necessary to park primarily motivated us to research a parking guidance system
the vehicle, and the need is increasing day by day as there that is reliable, efficient, and economical.
are more cars on the road than the number of parking spaces
[3]. The only way to optimize the use of parking space and B. Novelty and Contributions
reduce the parking wait time is to install a proper parking The proposed parking guidance system presents a unique
guidance system. For this reason, University of South Florida method of vehicle detection using an ultrasonic sensor. Al-
(USF) Center for Urban Transportation and Research (CUTR) though researchers have used ultrasonic sensor to detect empty
proposed a smart parking guidance system project to improve parking spaces, to best of our knowledge, no prior researcher
the parking experience for students, faculty, and visitors on have used it to detect vehicles in motion. We also show
campus. The project is funded by Student Green Energy Fund the logic involved in detecting the vehicle using flowcharts.

978-1-4673-9519-9/16/$31.00 ©2016 IEEE

 TABLE I
C HARACTERISTICS OF SENSORS

Sensor DC I P Accuracy Cost Range Area

(V) mA mW ($) (m) m2

Active 4.5 10 cm 0.2

Infrared to 33 165 to $14 to -
Sensor 5.5 30 cm 1.5
Passive Accurate
Infrared 4.5 0.1 2 within $15 0 41.9
Sensor to range to m2
(Standard 5.5 5
Fig. 1. Preliminary CO2 estimation by CUTR Detection
Type)
Passive Accurate
Infrared 3 0.17 0.57 within $15 0 2.8
The system is cheaper than any other available parking tech- Sensor to to range to m2
nologies. We built a working prototype and demonstrated its (Spot 6 0.85 5
efficacy in detecting vehicles reliably in an on-campus garage Detection
Type)
over 5 days. We survey extensively the possible sensors and Passive Accurate
compare them in detail (Table I). This is useful for future Infrared 0.17 0.85 within $15 0 19.6
researchers working on this problem. Sensor 5 to to range to m2
(Slight 0.3 1.5 2
Detection
II. R ELATED R ESEARCH Type)
Passive Accurate
Infrared 0.17 0.85 within $15 0 196
This section provides a study of sensor technologies used in Sensor 5 to to range to m2
existing parking locations. Then a literature review is presented (10 m 0.3 1.5 10
to find a reliable and efficient sensor that can be used for Detection
Type)
detecting a vehicle.
Accurate Varies
Microwave 4.75 30 15 within $10 on -
Sensor to to to set set
A. Study of Existing Parking Technologies 5.25 40 100 frequency freq.
High
The current sensor technologies active infrared, passive Ultrasonic 5 20 100 accuracy $2 0.02 1.1
infrared, microwave, inductive loop, ultrasonic and RFID Sensor up to to m2
are studied (from Table I) and compared with the proposed 3 mm 4.5
parking guidance system (Table II). The metrics considered for
comparison are installation, maintenance, cost and accuracy of TABLE II
the systems. C OMPARISON OF SENSOR TECHNOLOGIES
Inductive loops give high accuracy, but can be costly. RFID
technology has high precision, but it detects only cars with Sensor Installation Maintenance Cost Accuracy
tags. This may lead to an inaccurate count over time. Also, Technology
Active Infrared Ceiling Medium Expensive Accurate
the price of readers is very expensive. The parking count with Sensor Mount
infrared and ultrasonic sensor technologies is estimated by Passive Infrared Ceiling Low Expensive Unreliable
installing these sensors in each space. Installing sensors for Sensor (PIR) Mount
Microwave Ceiling Medium Expensive Accurate
each space is costly and requires maintenance. Also, the PIR Sensor Mount
sensor is affected by climatic conditions and is unreliable. Inductive Surface
Microwave sensor technology gives an accurate count, but is Loop Pavements/ Low Very Highly
Sensor Ceiling Expensive Accurate
very expensive. The proposed system is economical, requires Mount
low maintenance, easy to install and is accurate. The only Ultrasonic Ceiling Medium Expensive Accurate
disadvantage of the system is that it can incorrectly counts Sensor Mount
golf carts and peoples walking under the sensor as vehicles. RFID Ceiling Medium Expensive Accurate in
Mount controlled
Mount situations
B. Research of Sensor Technologies Proposed Ceiling Low Economical Accurate
System Mount
The selection of a sensor for this kind of application is a
challenging task. Sensor technology can be classified as visual
and non-visual. Visual based sensors include still and video 1) Visual Sensors: Visual based detection of a vehicle uses
cameras, while non-visual based sensors includes infrared, video cameras. Video image cameras use image processing
ultrasonic, and microwave sensors. and license plate recognition techniques to detect vehicles
[4]. Cameras installed at both the entrance and exit of the A. Block Diagram
parking lot are used to maintain an overall vehicle count.
Image processing techniques are used to detect a vehicle [5]. Figure 2 shows the block diagram of the parking guidance
A vacant spot is detected by installing a camera for three system. Arduino Uno, ultrasonic sensor, and temperature sen-
spots. License plate recognition identifies the vehicle and its sor are used to build the system. Ultrasonic sensor works
information [6]. Using this system on a large scale involves on the principle of sound waves and the speed of sound is
high installation costs. sensitive to air temperature. Temperature sensors estimates
Visual based detection methods have a drawback in accu- the speed of sound which removes the error due to changing
racy as they are easily affected by environmental conditions. weather conditions. First, ultrasonic and temperature sensors
The primary problems are occlusion effects, shadow effects [5] are connected to the Arduino microcontroller. The controller
and improper lighting conditions. A fully occupied space can detects the vehicle by triggering the sensor, and then it counts
be counted vacant in high lighting conditions, and an empty the vehicle and triggers the camera to send images to the web
space may be considered occupied in poor lighting conditions server.
due to shadow effects [5]. Further research has been done in The working model of a counter version is shown as simple
identifying a reliable sensor to detect a vehicle. count system in Figure 2. Here, the microcontroller checks
2) Non-Visual Sensors: A detailed study of sensors that can for obstacle by triggering the ultrasonic sensor. Whenever an
be mounted on the roof of the parking garage are considered. obstacle is detected, the controller determines if the obstacle is
Specific sensor technologies are active and passive infrared a vehicle or not by triggering more pulses. The logic involved
sensors, ultrasonic sensors, and microwave sensors. A com- in vehicle detection and count is thoroughly explained in
plete study on characteristics is done in thesis work [7]. Section III-D. Once the vehicle is detected, the web server
After a thorough review of characteristics, these sensors are is updated with increment/decrement in the number according
compared that helps in the proper selection of the sensor. to entry/exit as shown in dotted lines. For the overall count
The operating range of temperature of -10◦ C to +50◦ C of of the parking lot, this kind of system is installed at each
all the sensors The system requirements are 5 m range, 4- entry and exit and is integrated with the web server. The web
5 m2 of sensitive area and few dollars cost. Active infrared, server calculates the overall count by subtracting the number
passive infrared (PIR) and microwave sensors are ruled out of vehicles exited from the number of vehicles entered.
for our application for several reasons. Active infrared has The parking guidance system is designed to trigger an
a short range of detection of 1.5 m. Standard PIR has a external camera to take a video/image when a vehicle is
excessively wide sensing area of 41.92 m2 . Slight PIR has detected. Axis Q1614, a network camera is used in this system.
poor range of detection of 2 m. 10-meter detection PIR has It has four I/O ports in which one of them is trigger input.
large sensing area of 195.57 m2 , which is too high for this Once the processor detects a vehicle, it sends a pulse to the
application. Ideally we want the sensing area to be wide camera through one the I/O pin of the camera.
enough to accommodate a standard vehicle. Microwave sensor
requires external installation of an antenna that requires its
maintenance. Ultrasonic sensor stands out to be the most B. Vehicle Detection by Ultrasonic Sensor
reliable, efficient, and economical sensor. It has an acceptable
range of detection of 4.5 m and a safe operating range of Vehicle detection in this system is achieved with HC-SR04,
temperature of -10◦ C to +50◦ C. It is efficient because it has an ultrasonic sensor. This sensor offers an excellent non-
a fast response time of 25 milliseconds and sensing angle of contact range detection with high accuracy. HC-SR04 consists
15◦ . It is economical because it is available for less than $2 of a transmitter head that converts electric pulse to sound
and easy to install in the field. pulse, and a receiver head that converts sound pulse back to
an electric pulse. The transmitter head sends a sound pulse
and the receiver head receives it back if any obstacle is
III. P ROPOSED PARKING G UIDANCE S YSTEM
present within its range of detection. This sensor contains four
The proposed embedded system detects a vehicle, triggers pins: Vcc, GND, Trigger and Echo pins. Vcc and GND are
the camera to capture images and maintains the count of the connected to 5 Volts and ground respectively of the processor.
vehicles. The system uses an ultrasonic sensor, an Arduino The processor pulses the trigger pin and sends a sound pulse
microcontroller board, and a temperature sensor. The system throught the transmitter head. If any obstacle is present within
maintains the count of the vehicles, which provides the number the range of detection, the sound pulse hits the obstacle and is
of parking spots available in the parking lot. Also, a camera is reflected back to the receiver head. The receiver head receives
triggered to take 2-5 images the moment a vehicle is detected. the reflected sound pulse and sends it to the processor through
These images were used for further research and also to echo pin. The sound frequency of HC-SR04 is 40KHz, which
verify the system’s efficiency. The two versions of the system is above the human hearing frequency range (20 Hz to 20
are: counter version which maintains the overall count of the KHz). The working current ranges from 10 mA - 20 mA. The
garage and, trigger version which triggers camera for decal size of the sensor is small (45mm x 20mm x 15mm), which
recognition. This thesis reports the counter version. makes it easy for installation.
 Fig. 2. System Block Diagram

C. Arduino Microcontroller Board

Arduino UNO is the microcontroller board used in this D = 1/2 ∗ (v ∗ t) (2)
application. It has easy to use hardware for prototyping, low
cost and has simple software development interface. It contains The ultrasonic sensor is installed to the ceiling at en-
ATmega328 microcontroller chip with a clock speed of 16 trance/exit in a parking garage so that it detects the ground
MHz, which accelerates the system response time. ATmega328 as an obstacle. The distance between the sensor and ground
is a low power AVR 8-bit microcontroller with 2 KB SRAM. is considered as Maximum Range. When the system is
It has 23 programmable I/O lines of which Arduino UNO powered on, the processor pulses the sensor and calculates
provides 12 digital I/O pins (2 - 13) and 6 analog I/O pins the maximum range. After calculating the maximum range,
(A0 - A5). the processor pulses the sensor continuously and checks for
The trigger and echo pins of the ultrasonic sensor are a change in the range. When an obstacle enters/exits the
connected to the digital pins 2 and 3 of the board respectively. garage, the obstacle height is calculated by comparing obstacle
After the system is powered on, the processor sends a pulse, distance with maximum range. The obstacle height calculated
starts an internal timer, and waits for a given timeout. If is used for vehicle detection and is explained further in Section
an obstacle is detected within the timeout, the processor III-D1.
receives the pulse through the echo pin and stops the timer. 1) Vehicle Detection and Count: The flow chart of logic
The processor returns the time taken by pulse to travel in to detect and count vehicles using an ultrasonic sensor is
microseconds. This time is used to calculate the distance of shown in Figure 3. After initial field tests the minimum vehicle
the obstacle, as explained in Section III-D later. If no obstacle height is 50 cm, and considered to be the minimum threshold
is detected, the timer stops after the timeout and sends a pulse height. The controller calculates the temperature of the air
again. and calculates the speed of sound from it. Then it generates a
pulse and calculates the maximum range as discussed in above
D. Logic section. Next, the controller calculates the obstacle height
As discussed the controller pulses the ultrasonic sensor and by subtracting new obstacle distance from the maximum
calculates the time taken by the pulse to return. The distance range. The minimum height of the vehicle is set as minimum
of the obstacle is calculated using the speed of sound in air as threshold height, and the obstacle height is compared with it.
the ultrasonic sensor works on the principle of sound waves. If the obstacle height is greater than threshold height, then it
Equation 1 estimates the speed of sound (m/s), v, in dry air is considered to be an obstacle that can be a vehicle. This loop
at a given temperature, T , (in Celsius). repeats for every five milliseconds and checks for the presence
of an obstacle (vehicle).
v = 331.5 + (0.6 ∗ T ) (1) After detecting an obstacle, the count logic checks whether
it is a vehicle or not. A vehicle is a lengthier object, so time
As the speed of sound varies with the air temperature, taken by it to cross the sensor will be more than time taken by
a temperature sensor is used to correct the errors due to the processor to complete one loop. The processor is so fast
changing weather conditions. The distance calculated is twice that it can complete a loop within 5 milliseconds, as discussed
the actual distance as the pulse travels from the transmitter above for obstacle detection. For this reason, experiments
to an obstacle, bounces back from the obstacle, and finally are conducted to determine a minimum number of pulses
reaches the receiver. Equation 2 estimates the distance of a vehicle can take to cross the sensor, and minimum pulse
obstacle based on the estimated speed of sound (v), and time width threshold is calculated. These calibration experiments
elapsed (t). are discussed in Section IV. Figure 3 shows the logic employed
 discussed in Chapter 4.
E. Power Consumption
Power consumption of the system indicates its efficiency.
The power consumption is given by current multiplied by
the supply voltage. The supply voltage is 5V. Arduino con-
sumes 232.5 mW, ultrasonic sensor consumes 100 mW, and
temperature sensor consumes 0.25 mW. The overall power
consumption of the vehicle detection and count system is
332.75 mW.
F. System Cost
The proposed parking guidance is cost-efficient. TMP-36 is
the temperature sensor used, and IRF510 is the nMOSFET
used. The cost of Arduino Uno, ultrasonic sensor, temperature
sensor and IRF 510 are $25, $1, $1.50, and $1 respectively.
The overall cost of the system to detect and count the
number of vehicles entering the parking lot is less than $30 per
each entrance/exit. Triggering a camera uses the same system
used for vehicle detection and count. The cost of the system for
this is less than $30 per each entrance/exit. All these costs can
be still reduced by designing custom PCB with ATmega328
microcontroller for their application.
IV. E XPERIMENTAL R ESULTS
After a thorough study of the ultrasonic sensor, experiments
were conducted to test the sensor. Experimental setup required
was designed to test the sensor in the field. The sensor
was calibrated and initial feasibility tests were performed at
Richard Beard Garage at the University of South Florida
during weekends. Initial test data collected was by driving
cars at various speeds. These results were analyzed and logic
was developed for vehicle detection and count. Later, this
system along with Axis Q1614 camera was deployed on the
Fig. 3. Flow chart of the logic to count the number of vehicles west entrance of Richard Beard Garage to run several trials
to validate system. The Richard Beard Garage has a total of
3000 spots. It was determined that 5% to be an acceptable
margin of error with system reset every week.
for vehicle count. First, an obstacle is detected using the
detection logic. If no obstacle is detected, the processor stays A. Experimental Setup
in the inner loop to detect the obstacle. If any obstacle is A simple detection box is enough for data collection of
detected, then the processor enters the outer loop and checks vehicle detection as shown in Figure 4. The detector includes
the obstacle for minimum pulse width threshold by sending a an ultrasonic sensor, temperature sensor, connecting wires, and
set number of pulses. The processor exits out of this loop, if an Arduino microcontroller board. The sensor was connected
obstacle does not satisfy the condition specified. If the obstacle to an Arduino microcontroller board using connecting cables.
satisfies the required condition, i.e., it is present for minimum An external battery was used to power the system during
pulse width threshold, then processor determines that it is a the initial feasibility tests. A computer compiled code to the
vehicle and increments/decrements vehicle count. If vehicle microcontroller board. This detector is mounted temporarily
stays under the sensor for a longer time, the processor enters at a height of 2.4 meters. Vehicles driven exactly under the
the outer loop, waits until the vehicle leaves the sensor arena. detector yielded good results. The system was powered by
This logic is accurate which can record overall count of the supplying power lines from the camera.
parking garage.
The logic is same except that when a vehicle is detected, a B. Calibrating the Sensor
digital pin on Arduino is set high which is connected to the The ultrasonic sensor works on the principle of acoustics.
I/O pin of the Axis Q1614 camera. Upon receiving the pulse, The speed of sound varies with the temperature of the air. A
the camera can be configured to send a video/image to the temperature sensor (TMP-36) is used to estimate the speed of
web server. These images are used to validate the system as sound. Also, when the system is powered on, the sensor should
 Fig. 5. Plot of distances of slow moving vehicles (5 mph)

Fig. 6. Plot of distances of fast moving vehicles (20 mph)

Fig. 4. Parking System experimental setup

D. Field Trials
detect the ground surface first to calculate the maximum range. The detector along with Axis Q1614 camera was deployed
There should be no obstacle underneath the sensor during the at the West entrance of Richard Beard Garage as shown in
system startup. After calibrating the sensor, initial feasibility Figure 7 to test the working of the system. The I/O connector
tests were conducted. The test results are shown in [7]. of the camera has output power line on pin 2 which is 12V
DC. The power to the detector box was supplied from pin
C. Results from Initial Feasibility Tests 2 and pin 1 (GND). As discussed, Axis Q1614 is a network
After installation of the sensor for vehicle detection, cars camera that can send images from the camera to the database
at various speeds drove underneath the sensor, the distances when it is triggered. When a vehicle enters the garage, the
detected were analyzed, and variation of obstacle distances sensor system triggers the camera, and the camera would send
were calculated by the controller. ATmega328 triggers the the images to the local database set up at Center for Urban
pulse and calculates the echo time. The distance of the obstacle Transportation and Research (CUTR).
is calculated every 5 ms. When a vehicle approaches the After conducting trials, the number of cars detected by the
sensor, the distance of the obstacle changes and if we plot system for five days are shown in [7]. Table III shows the
those distances on a graph we can expect an inverted shape number of cars detected by the system for 24 hours during
of the car. The test results are shown in Figure 5 and Figure the first trial. “Exiting Cars” are cars exiting from the garage
6. The results attained from the first set of experiments to and detected as cars entering the garage as shown in Figure 9.
detect a vehicle using ultrasonic sensor were as expected. “P” are persons entering the garage and detected as cars. “Golf
Three cars drove underneath the sensor and distances detected Carts” are small vehicles as shown in Figure 8. Golf Carts are
were plotted with distance in meters as Y axis and time in considered as an error because they are not supposed to park
milliseconds as X axis. Figure 5 is the plot for the first trial and in the actual parking spaces. “Cars det.” are the total number
Figure 6 is the plot for the second trial. The curved portions of of vehicles identified as cars by the system. “Act. Cars” are
both plots indicate the presence of a car and flat region shows the cars actually entered the garage excluding the golf carts,
the absence of a car. exiting cars, and people. “E” is the Error accumulated. The
After analyzing the results acquired from preliminary tests, “Cars Detected” and “Act. Cars” detected for first trial are
the minimum time taken by a fast moving vehicle to cross compared and plotted against the “Hour” and shown in Figure
the sensor was less than 100 milliseconds (Figure 6). These 10. Similarly data was collected for four more days and shown
preliminary tests gave a minimum number of pulses a vehicle in Figure 11. Refer to thesis [7] for detailed results.
can encounter. This data was used to set a minimum number of Table IV shows the average error for five days. The total
pulses, i.e., threshold pulse width, for vehicle count. This can number of cars detected were 4,947 out of which 4,694 were
eliminate the error of counting a person as a vehicle, because actual cars. We can see that 70% of the error is due to exiting
the pulse width of a vehicle is greater compared to a person. cars. This is because the sensor is installed over the entrance
 Fig. 7. System installation at West entrance of Richard Beard Garage

TABLE III
DAY 1 - A PRIL 15 TO A PRIL 16, 2015

Hour Cars Act. Golf Exiting P E Percent

det. cars Carts Cars Error
1st 19 17 2 0 0 2 11.76
2nd 19 16 1 2 0 3 18.75
3rd 20 17 0 3 0 3 17.64
Fig. 8. Example of a golf cart entering the garage 4th 34 28 1 3 2 6 21.42
5th 26 23 1 2 0 3 13.04
6th 19 18 0 1 0 1 5.55
7th 16 13 0 3 0 3 23.07
8th 8 8 0 0 0 0 0
9th 8 8 0 0 0 0 0
10th 1 1 0 0 0 0 0
11th 1 1 0 0 0 0 0
12th 7 5 1 1 0 2 40
13th 31 31 0 0 0 0 0
14th 82 81 0 0 1 1 1.23
15th 167 166 0 0 1 1 0.60
16th 125 124 0 1 0 1 0.80
17th 150 148 0 2 0 2 1.35
Fig. 9. Example of cars exiting from the garage 18th 160 157 1 2 0 3 1.91
19th 101 98 1 2 0 3 3.06
20th 90 85 3 2 0 5 5.88
lane however the cars straddled over entrance/exit lanes. This 21st 65 62 1 2 0 3 4.83
error can be removed by separating the entry and exiting 22nd 54 46 0 5 3 8 17.39
23rd 69 63 0 6 0 6 9.52
pathways by a medium so that cars stay in their lanes. Further, 24th 33 32 0 1 0 1 3.12
if error due to Golf carts is not considered then the error is 24 hrs 1305 1248 12 38 7 57 4.56
further reduced as shown in Table V. If all these errors are
reduced then, the system reset can be done remotely every
month. size and low cost. Initially, the challenging task was to select
a reliable sensor for this application, which was overcome
V. C ONCLUSION AND F UTURE W ORK by a thorough review of sensor technologies. Compared to
The vehicle count feature monitors overall count of the other technologies, this system does not require installation
parking lot in real time. This system can help the drivers to of sensors into pavements and maintenance. Also, the system
know the available number of parking spaces before entering works in any weather condition. For this reason, it is suitable
the garage. The test results show that the system is accurate for all kinds of parking areas. In the future, the Arduino
enough to warrant further investigation. The proposed parking processor can be replaced by a PCB design using ATmega328
guidance system is easy to install because of sensor’s small processor that can further reduce the cost of the system.
 Fig. 10. Hour by hour comparison of estimated car count versus actual count.

is a vehicle or not in a particular parking spot using the

logic developed to detect a vehicle. This feature needs some
additional logic to glow red or green lights. The processor
glows red to indicate an occupied spot and green to indicate
an available spot. Additionally, number of available spots on
each floor can be known by installing the system at each
entrance/exit of each floor. If the system is installed at each
floor, it can give us number of cars that have exited and entered
each floor. This way, an overall count of each floor can be
maintained.

Fig. 11. Day by day comparison of estimated aggregate count versus actual ACKNOWLEDGMENT
count
This work has been supported and funded by the Student
Green Energy Fund (SGEF), Office of Sustainability (OoS)
TABLE IV
AVERAGE ERROR FOR FIVE DAYS at USF. The authors would like to thank Center for Urban
Transportation and Research (CUTR) and Parking and Trans-
Days Cars Act. Golf Exiting P E Percent
portation Services at USF for their help in this project.
det. cars Carts Cars Error
Day 1 1305 1248 12 38 7 57 4.56 R EFERENCES
Day 2 1213 1158 7 46 2 55 4.75 [1] E. Polycarpou, L. Lambrinos, and E. Protopapadakis. Smart parking
Day 3 1240 1185 11 34 10 55 4.64 solutions for urban areas. In World of Wireless, Mobile and Multimedia
Day 4 773 719 11 37 6 54 7.51 Networks (WoWMoM), 2013 IEEE 14th International Symposium and
Day 5 416 384 3 20 9 32 8.33 Workshops on a, pages 1–6, June 2013.
5 days 4947 4694 44 175 34 253 5.39 [2] Y. Geng and C.G. Cassandras. A new smart parking system infrastructure
and implementation. Procedia-Social and Behavioral Sciences, 54:1278–
1287, 2012.
TABLE V [3] R. Vishnubhotla, P.S. Rao, A. Ladha, S. Kadiyala, A. Narmada, B. Ro-
AVERAGE ERROR FOR FIVE DAYS WITHOUT GOLF CART ERROR nanki, and S. Illapakurthi. Zigbee based multi-level parking vacancy
monitoring system. In Electro/Information Technology (EIT), 2010 IEEE
International Conference on, pages 1–4, May 2010.
Days Cars Act. Exiting P E Percent [4] G. Revathi and V.R.S. Dhulipala. Smart parking systems and sensors: A
det. cars Cars Error survey. In Computing, Communication and Applications (ICCCA), 2012
Day 1 1305 1260 38 7 45 3.57 International Conference on, pages 1–5, Feb 2012.
Day 2 1213 1165 46 2 48 4.12 [5] A. Kianpisheh, N. Mustaffa, P. Limtrairut, and P. Keikhosrokiani. Smart
Day 3 1240 1196 34 10 44 3.67 parking system (sps) architecture using ultrasonic detector. International
Day 4 773 730 37 6 43 5.90 Journal of Software Engineering and Its Applications, 6(3):55–58, 2012.
Day 5 416 387 20 9 29 7.50 [6] M.Y. Aalsalem, W.Z. Khan, and K.M. Dhabbah. An automated vehicle
5 days 4947 4738 175 34 209 4.40 parking monitoring and management system using anpr cameras. In
Advanced Communication Technology (ICACT), 2015 17th International
Conference on, pages 706–710, July 2015.
[7] O. Dokur. Embedded system design of a real-time parking guidance
The vehicle space availability indicator can be an add- system. Master’s thesis, USF, 2015.
on feature to the system. It is simple to indicate if there