Evaluation of an Arduino-based IoT Person Counter

Bruno F. Carvalho, Caio C. M. Silva, Alessandra M. Silva, Fábio Buiati

and Rafael Timóteo

Laboratório LATITUDE – Faculdade de Tecnologia, Universidade de Brasília, Campus Darcy Ribeiro, Brasília, Brazil

Keywords: IoT Devices, Arduino, Motion Sensor, Ultrasonic Sensor.

Abstract: The IoT devices can provide a wide range of information, which can be used to infer the behavior patterns

with a large semantic bias. In this sense, an IoT network has the ability to use trafficked information to perform

its own management. One type of information that can be used by an IoT network is the amount of people in

a certain place. This information, combined with others, can help IoT-based systems discover characteristics

about the environment in which it is deployed. Thus, the integration of the data captured provides the

achievement of manifold applications, such as air conditioning regulation, security access and people

management in a working environment. In this research, it is proposed the implementation of an IoT person

counter. Two different technologies were used, aiming to verify the best option to design a counter device

with low-cost microcontrollers and sensors. Experimental results shows that in controlled environments the

IoT person counter has a satisfactory accuracy. Some limitations were also identified in order to clarify the

scenarios where those devices can be used.

1 INTRODUCTION

The Internet of Things enables opportunities for more

direct integration between the physical world and

computer-based systems. Devices inserted in an IoT

environment must be able to communicate and

propagate information about the digital and physical

environment, in which they are present. The

information gathered by those devices can be used to

provide services with fully awareness of current

execution environment, and also, can be used to

minimize the power consumption, processing time

and the amount of packages trafficked on the

network.

Information of several natures can be propagated

in an IoT network, from the device hardware

specification to the data inferred based on the

information captured from the physical environment.

For example, an IoT device can send a message with

a given information about the temperature, humidity,

luminosity, or motion detection, among others. Such

information can be used to characterize the execution

environment.

However, the information provided by IoT

devices have an innate characteristic of imperfection

and its quality is highly influenced by the way it is

gathered. In fact, this information may even be

incorrect. Most sensors feature an inherent inaccuracy

(e.g., a few meters for GPS positions). Furthermore,

each technology has its own specifications and

limitations, creating some technical restrictions

during the construction of an IoT device (Henricksen,

2004).

In this way, this paper presents the

implementation of an IoT device that has the ability

to inform how many persons there are in a room. This

information can be useful in a variety of scenarios: to

control the room temperature; to alert if the number

of persons exceeds the supported by the site; to send

information about trespassing; to count how many

persons are using public places, in addition to several

others.

Since imperfect or incomplete data may hinder the

proper functioning of an IoT-based system, the

proposal was implemented using two different

approaches: motion and ultrasonic technology. Those

technologies have been combined with

microcontrollers to create a device capable of

informing, with accuracy, the difference of

individuals entering and exiting a particular room. In

order to find the best alternative for this purpose, all

implementations were evaluated in the aspect of

validity of the information provided.

The remainder of this paper is organized as

follows. Section II presents a brief discussion about

IoT devices and used technologies. The IoT person

Carvalho, B., Silva, C., Silva, A., Buiati, F. and Timóteo, R.

Evaluation of an Arduino-based IoT Person Counter.

DOI: 10.5220/0005954601290136

In Proceedings of the International Conference on Internet of Things and Big Data (IoTBD 2016), pages 129-136

ISBN: 978-989-758-183-0

129

counter implementations are described in section III.

In section IV it is presented a set of experimental

results. Finally, conclusions and discussions of future

work are shown in section V.

2 RELATED CONCEPTS

In this section, we raise some concepts related to IoT

objects, in order to characterize the scenario, which

fits our proposal. First, on section 2.1 is presented

some concepts related to IoT devices. Next section,

describes concepts on microcontrollers and Arduinos.

Sections 2.3, 2.4 and 2.5 clarify the concepts related

to motion and ultrasonic technologies, respectively.

2.1 IoT Devices

The increasing number of smart devices, ranging

from computers to simple domestic appliances,

together with the ease of access to the internet,

brought up the notion of Internet of Things

(IoT). There are several ways to define IoT, but the

essence of the concept is the device interconnection.

IoT can be defined as a global network infrastructure

where physical devices, with a unique identity and a

virtual representation, have the ability to

communicate with other devices and distributed

architectures, like clusters, grids and clouds (Manna,

2014). Besides the communication with each other,

IoT devices are able to access information on the

internet, to retrieve and manipulate data and to

interact with users. Thus, it is possible to increasingly

observe the fusion of the physical and digital world.

In a typical IoT structure there are different types

of sensors, such as temperature, motion, humidity,

RFID, used to obtain specific data about the

environment. These sensors are connected with

embedded devices that collect and process the data

provided by the sensors (Toma, 2014).

2.2 Microcontroller

Primarily, it was called a microcomputer due to the

fact that includes built-in RAM, ROM, and I/O

(Gridling, 2006).

Differently from microprocessors, the objective in

developing microcontrollers was to create a complete

computer in a single chip with memory, peripherals

components and a processor that can be used as an

embedded system (Sickle, 2001). Nowadays, there

are embedded systems that require a minimal of

memory and processing, while others are extremely

sophisticated.

2.2.1 Arduino

Arduino is an open-source prototyping platform

composed by a microcontroller or physical

programmable circuit board and a software or IDE

that runs in a computer. The Arduino boards interacts

with the environment through peripheral components,

like electronic sensors, lights, and motors. Then,

using the Arduino programming language and the

Arduino Software (IDE), it is possible to manipulate

data captured (W. Arduino, 2016).

Nowadays, there are several different types of

controllers on Arduino platform, which differ mostly

on IO pins and version of flash memory (Raspaile,

2013). The most important ones are Arduino Uno,

Arduino Mega2560, Arduino Nano, Arduino Mini,

Arduino Due, Arduino Leonardo and Arduino ADK.

2.3 Motion Sensor

A motion sensor is a device used to detect movements

in a certain area and it can use multiple types of

technologies to perform this task, which defines

aspects such as range, precision and sensitivity

(M.Sensor, 2016). There are two main types of

motion sensors:

• Passive sensor: it detects variations in energy in

the surrounding area, but it does not emity

energy.

• Active sensor: it transmits infrared light,

microwave radiation, or sound waves and waits

for its response. Due to continuous activity, this

type of sensor consumes a significantly larger

amount of energy.

There are several applications for motion sensors.

The most common are: security systems, to open and

close automatic doors and to turn on lights when a

person enters a room.

2.4 Ultrasonic Sensor

The operation of an ultrasonic sensor is based on the

transmission of ultrasonic pulses and the time

response of the pulses. In this way, the principle

underlying this technology is that speed of sound in

air is approximately constant (Wijk, 1998). Thus,

estimating the time for pulse reflection allows

knowing the distance to the object due to the

proportionality relation. Therefore, ultrasonic sensors

are used frequently to detect objects and for distance

measurement applications.

These sensors are capable of detecting any type of

object that have sufficient acoustic reflectivity. On

IoTBD 2016 - International Conference on Internet of Things and Big Data

130

Table 1: Person counter REST request.

Method

Request Description

PUT http://server-ip/service?serv_name=A&dev_id=B&info=C

Generic request

PUT http://172.2.2.10/service?serv_name=count&dev_id=5&pc=5

Example of a request where the device id

is 5, the number of people in the room is 5,

and service name is count

PUT http://172.2.2.10/service?serv_name=count&dev_id=1&pc=9

Example of a request where the device id

is 1, the number of people in the room is 9,

and service name is count

the other hand, sound absorbing material, such as

cloth and foam, are not easily identified.

3 IoT PERSON COUNTER

In this section are presented the proposed IoT person

counters. Such devices have the ability to inform how

many people are present in the environment in which

they are installed. The IoT person counter takes

advantage of well-established technologies, for

example ultrasonic, to perform the counting of

individuals transiting in a certain place.

It is noteworthy that all these devices communicate

with a central server through a REST interface (Silva,

2016), informing their unique identification (dev_id),

the service name (serv_name) and the people count

information (pc). The data is sent via Wi-Fi technology

using an Arduino Wi-Fi shield (A.WiFi, 2016). Table

1 presents the generic structure for REST, as well,

examples of the requests performed by the IoT devices.

The proposal presentation is divided into two

subsections. First, it is presented the IoT person

counter based on motion sensor technology. Next

subsection, presents an implementation using

ultrasonic sensors. All subsections presents a circuit

schematic, an algorithm that performs the counting

process and the specification of technologies used to

build the devices. It also elucidates the limitations of

each technology and their impact on the capture of

environmental data.

3.1 Motion Approach

The motion approach for an IoT person counter was

designed with a microcontroller and two motion

sensors. The microcontroller verifies which sensor

was activated and waits for the activation of the other

sensor. If both were activated in a given time interval,

it is identified that a person has passed in front of the

device. The prototype of this device was built with an

Arduino Uno (A.Uno, 2016) and two motion sensors

model DYP-ME003 (Motion, 2011).

3.1.1 Circuit Schema

The circuit schema of the motion person counter is

presented on Figure 1. This circuit is composed by

two PIR sensors (motion sensor) connected with one

Arduino Uno. The power of these sensors (VCC) is

supplied by the 5V Arduino pin. The ground (GND)

of each sensor is attached in the GND Arduino pin,

the signal output of the first sensor is connected at the

digital pin 7 and the signal of the second one at digital

pin 8.

Figure 1: Motion person counter circuit schema.

In its default state, the motion sensor sends

through the signal pin a LOW (0) output, meaning

that there is no movement. When the PIR sensor

detects motion, the signal output is HIGH (1) on the

correspondent Arduino digital pin. The combine of

Evaluation of an Arduino-based IoT Person Counter

131

two of these sensors allows four different states: two

sensors on 0, one sensor on 1 and another on 0 or two

sensors on 1. When the two sensors are HIGH, can be

detected that a person passed in front of the sensors.

3.1.2 Algorithm

The motion approach algorithm is presented in

Algorithm 1. The implementation defines a procedure

named read_opposite_sensor(byte, bool, int), and

other named main. The read_apposite_sensor

receives as parameter: a byte, which indicates the

sensor_id, a boolean value that gives the direction in

which the person is passing and an int value that

represents the time interval when the sensor has to be

read.

This procedure performs the reading of the

opposite sensor during the time interval max_time, if

it recognizes any motion, then it detected passage of

an individual. Depending on the direction, the

procedure can add or subtract the value in the persons

counter.

The main procedure is executed while the device

is on and checks if any motion was detected in any of

the sensors. In the case of motion detection,

read_opposite_sensor function is called to see if

anyone passed in front of the other sensor. It is

important to notice that this implementation does not

use external libraries, just Arduino pre-defined

functions.

Algorithm 1: Motion person counter

procedure read_opposite_sensor(byte

sensor, bool direction, int max_time) :

void

int time = 0;

while(millis() - time <

max_time)

if(is_high(sensor) &&

direction)

number_of_people++;

else if(is_high(sensor) &&

!direction)

number_of_people--;

end while

time - millis();

end procedure

procedure main() : void

int number_of_people = 0;

while(device_on)

if(sensor_is_high(first))

read_oposite_sensor(second, 1);

else

if(sensor_is_high(second))

read_oposite_sensor

(first,0);

end while

end procedure

The motion sensors used on this implementation have

some technical restrictions that should be taken into

account as they directly impact in the device

implementation: The sensor stands in HIGH for 2.5

seconds: When the DYP-ME003 sensor detects

motion, it stands in high for approximately 2.5

seconds. In this time interval, the sensor is

incapacitated to perform data capture. This limitation

undermines the counting process, if people pass in a

time interval shorter than 2.5 seconds. Besides that,

only one person can pass in front of the device, if a

group of people pass, the device will recognize as just

one person.

3.2 Ultrasonic Approach

The ultrasonic IoT person counter was built with one

microcontroller and two ultrasonic sensors. This

device performs its functionality similarly to the

motion device. The difference consists on the

information provided by the sensor. The ultrasonic

device informs the distance of an object that is ahead

of it.

Therefore, the device checks at what distance an

object is from the sensor and waits until the other

sensor recognizes the object in the same distance

interval. In this case, it is recognized that someone has

passed in front of the device. The prototype of this

device was built with an Arduino Uno (A.Uno, 2016)

and two ultrasonic sensors model HC-SR04

(Ultrasonic, 2011).

3.2.1 Circuit Schema

The ultrasonic circuit schema is shown in figure 2. An

Arduino UNO is responsible for collecting the

information captured by the ultrasonic sensors. This

circuit has two HC-SR04 sensors (ultrasonic sensor)

connected with one Arduino Uno. The VCC is

supplied by the 5V Arduino pin and the GND of each

sensor is attached in the GND Arduino pin. Each

sensor has one Trigger and Echo pin. The first sensor

has Arduino digital pin 7 and 6 connected,

respectively, to Trigger and Echo. Besides that, the

second one is attached with digital pin 9 on Trigger

and digital pin 8 on Echo.

The ultrasonic sensor constantly measures the

distance of objects ahead, sending a signal through

Trigger pin e receiving through Echo. When it is first

turned on, the device measures the distance of all

stationary objects in the environment. In this paper,

IoTBD 2016 - International Conference on Internet of Things and Big Data

132

this distance will be called environment distance.

Besides that, when a person passes in front of the

device, the distance measured will be named person

distance.

The combining of these two sensors heads to four

different states: two sensors reading the environment

distance, one reading environment and the other a

person distance, and the two sensors detecting a

person distance. When two devices read a person

distance, that means someone passed in front of it.

Figure 2: Ultrasonic person counter circuit schema.

3.2.2 Algorithm

To perform people counting process the algorithm

presented in Algorithm 2 was deployed in the circuit.

It use two procedures and one library to compute the

information. The library is called “ultrasonic.h” (),

and it is used to perform two types of processing:

• timing(): Gives the time passed between the

emission and reception of an ultrasonic wave.

• convert(bool, int): receives the sensor id (in this

case a boolean value), and the precision wanted

(in this case centimeters).

The library was adapted to work with boolean

values as a sensor id. Besides that, two procedures

were implemented; read_ultrasonic_distance(bool)

and count_persons(float,float,bool). The first one

uses the library functions to return the distance that an

object is from the sensor. The count_persons

procedure counts the number of persons who passed

in front of one sensor.

The main procedure uses the following pre-

configured information: delay, min_dist and

max_dist. Delay indicates the time when the

microcontroller has to switch the direction of reading.

For example, if the delay time is set to 200

milliseconds, the device will verify how many

persons are entering the room for 200ms, then it will

verify how many persons are exiting the room for

200ms. The min_dist and max_dist represents the

distance range, this interval indicates the size (width)

of the corridor where people can pass.

Therefore, the algorithm is executed while the

device is on, and checks if anybody has passed in any

direction. If so, this amount is incremented or

decremented from the people global counter. In each

iteration the direction is changed so the device can

read in both directions. To finish the procedure the

device sends to the server the captured information,

through the rest interface, in order to preserve a

historical ballast.

Algorithm 2: Ultrasonic Persons Counter

procedure read_ultrasonic_distance(bool

direction) : float distance

microsec = ultrasonic.timing();

distance =

ultrasonic.convert(direction,

Ultrasonic::CM);

return distance;

end procedure

procedure count_persons(float

min_distance, float max_distance, bool

direction): int amount_people

float distance_dir = 0;

float distance_oposite_dir = 0;

number_of_people = 0;

distance_dir =

read_ultrasonic_distance(direction);

distance_oposite_dir =

read_ultrasonic_distance(!direct

ion);

if(same_distance_range(dis

tance_dir,

distance_oposite_dir))

number_of_people++;

while(distance > min_distance &&

distance < max_distance)

return number_of_people;

end procedure

procedure main() : void

float delay = 200;

int max_dist = 100; //centimeters

int min_dist = 40; //centimeters

int direction = 1;

amount_of_people = 0;

Evaluation of an Arduino-based IoT Person Counter

133

while(device_on)

while(time < delay)

if(direction)

amount_of_people +=

count_persons(min_dist,

max_dist,direction);

else

amount_of_people -=

count_persons(min_dist,

max_dist,direction);

end while

direction = !direction;

send_to_server(amount_of_people);

end procedure

The ultrasonic counter has limitations related to

the distance that the object is from the device. In other

words, if someone passes in front of the device at a

distance greater than one meter it usually loses

precision.

In addition to that, if there is a group of people

passing in front of the device, in the same distance

range, it is possible that the proposed device also

loses accuracy. Even with those limitations, for

counting a large numbers of people the ultrasonic

approach was more consistent than motion

technology based approach.

4 EXPERIMENTAL RESULTS

4.1 Experimental Environment

In order to compare the proposed solutions in a real

scenario, all IoT person counters were installed in the

same laboratory room. The devices were deployed

next to the door, objectifying only to capture data

related to the entry and exit of people. This room is

used by fifteen distinct persons at different times of

day.

The proposed implementations were analyzed in

relation of the accuracy of the information provided

by the IoT devices. Experiments were conducted in

the following approach:

• Controlled tests: Devices were tested in specific

scenarios with minimum external interference.

These scenarios can be seen in Figures 3,4,

5,6,7,8 and 9.

The scenarios proposed for the execution of controlled

tests were designed in order to analyse the

effectiveness of implementations in different

situations. Thus, the scenarios differ in number of

people, time interval between the passage of persons

and direction of passage, in other words, entering or

leaving.

Figure 3: Scenario 1 - One person entering or leaving the

room.

Figure 4: Scenario 2 - One person entering and other person

leaving the room at the same time.

Figure 5: Scenario 3 - Two persons entering or leaving the

room at the same time.

Figure 6: Scenario 4 - Two persons entering or leaving the

room in a short time interval.

IoTBD 2016 - International Conference on Internet of Things and Big Data

134

Figure 7: Scenario 5 - One person entering and other person

leaving the room in a short time interval.

Figure 8: Scenario 6 - A group of N people entering and

another group of M people leaving the room in a short time

interval.

Figure 9: Scenario 7 - A group of N people entering and

another group of M people leaving the room, all at the same

time.

Table 2 presents the result of accuracy evaluation

for all controlled tests. Each row represent a specific

scenario evaluated by motion and ultrassonic

approaches.

For scenarios where only one individual passes in

front of the device, in a controlled time frequency, the

devices obtained results next to those expected,

indicating that they can be used in real environments.

Both motion and ultrasonic technologies have a

inactivation time range limitation, respectively, of 2,5

seconds and 500 milliseconds.

The experimental results also indicate that in

scenarios with heavy human traffic both

implementations have an undermining accuracy.

Besides that, in scenarios where there are several

persons passing at the same time the two approaches

recognize as one individual.

Table 2: Accuracy of proposed implementations.

Scenario

Accuracy Result:

Ultrasonic

Accuracy Result:

Motion

1 90% 81%

2 0% 0%

3 0% 0%

4 87% 50.3%

5 63.2% 49.8%

6 58.7% 12.4%

7 0% 0%

5 CONCLUSIONS AND FUTURE

WORK

The research work presented in this paper proposes a

IoT person counter implemented with two different

technologies, motion and ultrasonic. The information

provided by person counter device can be used in

several scenarios, such as to ensure security of access

to a room, or to regulate the temperature of a room.

Algorithms have been suggested to carry out the

management of the information provided by the

sensors. As result of the proposal, experiments

indicate that in controlled environments, where there

are not many people transiting, both approaches have

a considerable accuracy. In contrast, scenarios in

which the flux of people is significant, both

applications have an impoverishment of its accuracy.

Given this scenario, where devices are

susceptible to imprecision of the information

captured from the environment, it is understood that

it is not possible to send information, continuously

and accurately, over a given situation occurring in the

environment, with only one IoT device. In this sense,

for future work it is proposed an IoT network person

counter (INPC). The INPC utilizes information

provided by different devices on the network to infer

how many people are in a certain place. Such

implementation will take advantage of information

provided by an IoT authenticator (based on RFID

technology), an IoT person counter (based on motion

technology), and an IoT face recognizer (based on

face recognition algorithms). Those three devices will

send information to the network and all the

information combined tends to indicate with a

Evaluation of an Arduino-based IoT Person Counter

135

better accuracy and precision the amount of the

number of people in a certain place.

ACKNOWLEDGEMENTS

The authors wish to thank the Brazilian research,

development and innovation Agencies CAPES

(Grant FORTE 23038.007604/2014-69),FINEP

(Grant RENASIC/PROTO 01.12.0555.00) and

PNPD/CAPES - Programa Nacional de Pos-

Doutorado/CAPES for their support to this work.

REFERENCES

A.Uno, Arduino Uno datasheet. On-line reference acessed

in 19/02/2016. https://www.arduino.cc/en/Main/

ArduinoBoardUno

A.WiFi, Arduino WiFi shield datasheet. On-line reference

acessed in 17/02/2016. https://www.arduino.cc/en/

Main/ArduinoWiFiShield

Gridling G., Weiss B. (2006). Introduction to

Microcontrollers. Vienna University of Technology.

Henricksen, K., Indulska, J. (2004). “Modelling and using

imperfect context information,” in Pervasive Computing

and Communications Workshops. Proceedings of the

Second IEEE Annual Conference on. IEEE, 2004, pp.

33–37.

Manna, S., Bhunia, S.S., Mukherjee, N. (2014). Vehicular

Pollution Monitoring Using IoT. IEEE International

Conference on Recent Advances and Innovations in

Enineering.

Motion sensor: spcification of DYP-ME003. On-line

reference acessed in

19/02/2016. http://elecfreaks.com/ store/download/

datasheet/sensor/DYP-ME003/ Specification.pdf

M. Sensor, The Beginner’s Guide to Motion Sensors. On-

line reference acessed in 23/02/2016. http://www.

safewise.com/resources/motion-sensor-guide

Raspaile, P., Keswani, V. (2013). An Approach Towards

Low Cost Wireless Sensor Network for the

Applications of IoT. International Journal of

Engineering Science and Advanced Technology.

Sickle, T. V. (2001). Programming Microcontrollers in C.

LLH Technology Publishing, Eagle Rock.

Silva, C. C .M., Ferreira, H. G. C., Sousa Júnior, R. T.,

Buiati, F. and Villalba, L.J.G, 2016. Design and

Evaluation of a Services Interface for the Internet of

Things. Wireless Personal Communications. p. 1-38.

Springer.

Toma, C., Popa, M. (2014). Internet of Thing Architecture

for Context Aware Sensors Data Processing in Waste

Management Solution. Journal od Mobile, Embedded

and Distributed Systems, vol. VI.

Ultrasonic sensor: specification of HC-SR04. On-line

reference acessed in 19/02/2016.

http://www.elecfreaks. com/store/download/product/

Sensor/HC-SR04/HC-SR04_Ultrasonic_Module_

User_Guide.pdf

Wijk, O., Jensfelt, P., Christensen, H.I., (1998).

Triangulation based fusion of ultrasonic sensor data.

IEEE International Conference on Robotic and

Automation.

W.Arduino, What is Arduino. On-line reference acessed in

19/02/2016. https://www.arduino.cc/en/Guide/

Introduction

IoTBD 2016 - International Conference on Internet of Things and Big Data

136