Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 11

Motorcycle Rider Helmet Detection Using Image

Processing
Juan Carlo Ocampo#1, Karl Vince Jimenez#2, Jambertvill Valerio#3, Jhon Manuel Ocampo#4, Marck Joshua
Punzalan#5, Aldrin Broqueza#6, Jayson De Dios#7, Engr. Asil Kastle Dela Cruz, MIT, PCpE#8
#
College of Computing Studies, Don Honorio Ventura State University
Cabambangan, Bacolor, Pampanga, Philippines
2018008280@dhvsu.edu.ph
1

2018003797@dhvsu.edu.ph
2

2015147685@dhvsu.edu.ph
3

2018003825@dhvsu.edu.ph
4

2018003666@dhvsu.edu.ph
5

2018003741@dhvsu.edu.ph
6

2018003378@dhvsu.edu.ph
7

8
aksdelacruz@dhvsu.edu.ph

Abstract— Wearing a helmet is a significant part of driving a I. Introduction


motorcycle, and most countries have strict helmet
regulations. Despite laws on helmet usage, there are still two- Helmet laws are administered to lessen head injuries and
wheeler riders who do not follow these policies. This study avoid fatalities. According to [1], helmet-wearing riders
intends to aid authorities by producing a system that assists that got involved in accidents had decreased the risk of
the users in capturing no-helmet riders by sending head injury by 69% along with a 42% decreased risk of
notifications through webpage and text messages. The death. This suggests that the majority of fatal head injuries
researchers utilized Raspberry Pi as the microcontroller, sustained in motorcycle collisions can be mitigated or even
accompanied by a GSM module for sending text messages, a prevented through helmet usage. However, despite the
web socket, a database, and an IP Camera for the visual traffic laws and years of helmets being subject to
input. In this study, Python programming language, HTML, continuous improvements in terms of safety and lifesaving
CSS, JavaScript, and PHP were essential in developing the qualities, as the use of helmets among riders keeps
system. The model was trained using collected images as declining, the numbers of motorcyclist fatalities keep
sample datasets through OpenCV. The system detects and
rising. Although there is an act in the Philippines, the
captures riders with a rate of 76.09% on accuracy tests
Republic Act No. 10054, that requires motorcycle riders to
conducted in three different barangays. The system provides
use standard protective motorcycle helmets while driving,
notifications through a webpage via the internet and sends
the usage of motorcycle helmets is still low due to a lack of
text messages with an average response time of 5.4 seconds
police manpower. Moreover, despite the penalties being
and 12.18 seconds respectively. The purposive sampling
technique was practiced to select 10 participants as end-user provided to violators, there still are people who do not
respondents to evaluate the system and three professionals to wear helmets as even though police officials tried to
assess the system through survey questionnaires using regulate this problem manually through checkpoint
ISO/IEC 25010 standards. The end-users considered the operations, not all of the violators can be monitored,
device as “Good”, with a calculated mean of 3.1 and the caught, and punished by the law. [2]
evaluation by the experts weighted an average mean of 4.18, This study focused on this problem and proposed a system
implying that the professional respondents are “Satisfied”. that will help law enforcers in improving the monitoring
These show that the system can serve the intended purpose system of checkpoint operations through the aid of
and that the overall function of the helmet detection system is developing a device that will detect and recognize non-
effective. helmet-wearing motorcycle riders and then notify its user
using Image Processing as the main technology to be used
Keywords— Motorcycle, Helmet, Detection, Notification, and the Raspberry Pi as the Microcontroller Unit.
Raspberry Pi, GSM, IP Camera, Python, JavaScript, PHP,
HTML, CSS, OpenCV, datasets, webpage.
According to [3], image processing refers to a variety of The captured images are stored locally and in a database as
operations that examine pictures or transform one image backup data to provide evidence for traffic law violations
representation into another. during checkpoint operations. The collected images are
The researchers then opted to run the system on a accessible by authorized personnel only through a user-
Raspberry Pi as the main processing unit, a credit-card- friendly webpage.
sized computer that connects to a computer monitor or TV With regard to the limitation of the study, the system relies
and works on a standard keyboard and mouse. It is a on electricity as a self-sufficient energy source is not part
powerful small-sized device that allows users to of the study and the detection of ICC-certified stickers on
communicate with computers and learn to write in helmets is not included in the system due to hardware
languages like Scratch and Python. restraints. Detection of plate numbers is not included.
A similar study entitled “Machine vision techniques for The product of this study would be beneficial to
motorcycle safety helmet detection” presented a system motorcycle riders, checkpoint personnel, Local
that detects motorcycle riders and determines if they are Government Units, and Future Researchers. This will
wearing safety helmets or not. The researchers of this allow motorcycle riders to be wary of not wearing a helmet
study utilized K-Nearest Neighbor (KNN) classifier. By violation when driving, for their own safety. The system
using this classifier, the device categorized motorcycles will enable the checkpoint personnel to keep track and be
and other moving entities based on the features extracted more aware of the violators that drove past them. The
from their region properties. Using KNN, the system device intends to help the Local Government Units to
determines whether the rider is wearing a helmet or not impose the law in a more efficient way with regard to
based on attributes extracted from four sections of the wearing helmets when driving a motorcycle. Upon the
segmented head area. The results of the experiments completion of the study, it will serve as a supporting study
demonstrate that the average correct detection rate for the for future researchers who are looking for ideas or topics
near lane, distant lane, and both lanes is 84%, 68%, and related to this study or for those who intend to further
74%, respectively. [4] develop, enhance, and innovate the product.
Correspondingly, [5] created an algorithm for the
automatic registration of motorcycle helmet wear using II. Methods
video data. The system was trained to recognize active
motorbikes, the number and location of riders on the The researchers employed the descriptive method as the
motorcycle, and helmet usage using 91,000 annotated methodology of this study. The researchers decided to use
frames of video footage gathered at several observation this method as it fits in the collection of the needed
sites around Myanmar. With a small amount of training for information for this study. [7]
individual observation locations, the system detects The first stage of development is the planning phase. The
motorcycle helmet usage with rates of 4.4% accuracy and a researchers discussed the design of the system to achieve
+2.1% accuracy when compared to a human observer. its intended purpose effectively. The proponents
With these supporting studies in mind, the researchers determined the components to acquire for the development
intend to help the authorities by developing a system that of the system. The researchers utilized Raspberry Pi 4
captures images and sends notifications to the user and that model which is the main controller unit. The Raspberry Pi
allows them to improve the strategies for capturing is also responsible for running the program. IP camera
disobedient motorcycle riders. By amplifying the methods serves as the visual sensor of the system that captures
of implementing the strict helmet law on two-wheeler target entities. The proponents decided to use the internet
riders, the system helps the authorities by providing as a medium for notification and a GSM module to enable
alertness and assistance to intercept violators due to the device for sending a text message as a secondary
driving a motorcycle without head protection is dangerous. notification method.
The proponents want to help mitigate fatal head injuries as
90 percent of deaths caused by motorcycle accidents were
riders that are not wearing helmets [6]. The researchers
also want to learn how image processing can contribute to
making the community safer while trying to minimize the
injury sustained because of accidents.
The study aims to produce a detection system for
motorcycle helmets using image processing with
notifications. The system may help notify its user after it Figur
captures images of riders who are not wearing a helmet. e 1: Block Diagram
Figure 1 shows the block diagram of the study. The motorcycle riders wearing and not wearing helmets. The
researchers utilized the use of Raspberry Pi 4 as the images are separated in two folders: negative and positive.
Microcontroller. The power supply will power the
Raspberry Pi and the IP camera, the microcontroller then
will supply the GSM module. The IP Camera will be used
to detect violations and the Raspberry Pi will send
notifications to the webpage and store images in the
webserver, another notification will be sent by a text
message that includes a link to the image capture through
the GSM module. The captured data is saved locally in the
Raspberry Pi and to the database for backup purposes
Figure 3: Annotation of positive images

Figure 4: Collected positive and negative images


2.2 TRAINING OF MODEL
Figure 2: Flowchart With the collection of datasets, the researchers used
Figure 2 shows the flow of the process of the system. The OpenCV to create Haar cascades and train the program.
process starts with the execution of the helmet detection The Haar Cascade is an object identification technique that
function, then the camera is in waiting state to detect a can recognize customized items in images or real-time
motorcycle. If the camera has detected a motorcycle, then video.
the action will proceed to the decision stage, the
differentiation between a positive and negative entity. If
the detected rider is wearing a helmet, the task will return
to the motorcycle detection state, however, if the detected
rider is not wearing a helmet, the activity will proceed to
the capturing of image. Right after the caption of the
subject, the system will save the image in the local storage
of the raspberry pi and to the database. Then the process
will proceed to the sending of notifications through the
GSM module and through the web.
2.1 COLLECTION OF DATASETS
After the complete concept of the system was finalized, the
software development for the helmet detection was started Figure 5: Vector File Creation and Training of Haar
by initializing the training phase of the model. The Cascade
researchers gathered sample data through capturing images The training of the Haar cascade was initialized after the
of both positive and negative entities on roadways. annotation and the creation of the vector file shown in
The proponents utilized 1000 positive images and 2000 Figure 5. The XML files are the output model that
negative images for the training of the model as shown in differentiates motorcycle riders with no helmets from
figure 3. The manually collected images are composed of motorcycle riders wearing helmets.
images that were captured by the proponents of themselves 2.3 SYSTEM DEVELOPMENT
while riding a motorcycle with and without a helmet and Following the completion of the training phase, the
captured images of passing motorcycles through roadways. researchers tested the model virtually. The virtual testing
The images collected from the internet came from websites used prerecorded videos to check if the detection model is
such as [8], [9], [10], [11] which contain photos of working.
Figure 6 shows the initial testing of the model as it was
tested internally using a laptop and a premade video. The
proponents determined that the detection model is working
as intended, the researchers imported the code to the
raspberry pi and the hardware development phase was
started by the assembly of the connection between the IP
camera and Raspberry Pi through the router. The device Figure 9: Webpage notification user interface on
was implemented on a post and checked the detection desktop browser and in Mobile browser
capability of motorcycle riders.

Figure 10: SMS Notification


The testing of webpage notification was conducted
Figure 6: Virtual Testing successfully, the proponents integrated the GSM module to
the Raspberry Pi and procured a SIM card that uses 2G
connectivity to function. A text message will be sent after a
few seconds to the registered number. The message
content says that a violator has been detected which serves
as an alert.
Following the trial testing of the web page and the text
message notifications completed, the prototyping phase
started with the assembly of the software and the hardware
Figure 7: Actual Testing at Barngay Saguin
of the system. The proponents added a script that allows
Figure 7 shows the actual testing of the Helmet detection
the program of the system to run on boot up of the
program. It was tested on an intersection inside Barangay
raspberry pi, creating the Motorcycle Helmet Detection
Saguin, City of San Fernando, Pampanga.
System.
Over the succeeding functionality assessments, the system
evaluation phase was conducted. The study aims to get
descriptive information from the end-users of the device:
the authorities at road checkpoints. The end-user and
respondents of this study are 10 officers who are involved
in road checkpoints. Moreover, three professionals from
Fig the same field of study are chosen to evaluate the product
ure 8: Raspberry Pi vs Laptop based on the International Organization for Standardization
The researchers ran the program on both devices: (ISO) and the International Electrotechnical Commission
Raspberry Pi and a Laptop. Figure 8 shows that the FPS (IEC) 25010.
(Frame Per Second) on the Raspberry Pi is only 0.24 and The selection of respondents is done through the purposive
the FPS on the Laptop is 1.96. This implies that better sampling technique. The data collection method applied in
hardware can run the program faster and more efficiently. this study was the questionnaires. Three professionals
As for the notification function of the system, the validated the questionnaire after they examined the
researchers then developed a webpage where notifications reliability of the questions and their content.
can be received and viewed when a violator is detected and
the use of a web database as the storage of the collected
images of the violators online. HTML, PHP, and
JavaScript are used to complete the web-based system.
Table 1. Assessment on the Ease of Operation of III. Results and Discussions
the Product

Figure 10. Motorcycle Riders Helmet Detection


System
Table 2. Effectiveness and Accuracy of the System
Figure 10 shows the whole system of Motorcycle
Rider Helmet Detection. This system is designed to
detect motorcycle riders who violate the R.A. 10054.
The device detects riders who do not wear helmets. It
will save the image of the violator on the webpage for
remote access and it is stored in the database and
The 4-point Likert Scale is administered to determine
locally in the raspberry pi for protection against
the assessment of the survey questionnaire for the 10
accidental loss of data. It will also send a notification
end users using the following range conversion shown
on the webpage and through text message indicating
in Table 3.
that a no-helmet motorcycle rider is detected. The
Table 3. Effectiveness and Accuracy of the System
notification in the text message also provides a link to
view the image. The hardware used are the following:
The Raspberry Pi was used as the main controller of
the system. It runs the detection program and sends the
Table 4. Evaluation questionnaire for experts
notifications to both SMS and Web channels. The
proponents opted for the version 4 8GB variant for
wider usage capabilities than the lower variants. The
camera used for the detection process is the TP-Link
Tapo C320WS. It has a 4MP definition with a 1080P
HD Resolution. It is wirelessly connected to the
Raspberry Pi through the router. A router was used to
connect two or more packet-switched networks or
subnetworks. In this study, a router was used as an
access point for both the Raspberry Pi and the IP
Camera while also providing them with an internet
connection.
Furthermore, three Computer Engineers, as
respondents, virtually evaluated the product based on
ISO/IEC 25010 evaluation questionnaire. To get the
overall rating of the quality of the product, the
conversion used for determining functional suitability,
performance, usability, reliability, portability, and
security categories in the ISO/IEC 25010 to determine
their satisfaction with the product is as follows: Figure 11. Website Contents
Table 5. Effectiveness and Accuracy of the System The content of the website includes the images saved
during an operation. The pictures are labeled with the
time and date of detection for better documentation
purposes. There are functions that can be used:
“Delete All” and “Export Images”. The “Delete All”
function deletes all the current images collected. The
“Export Images” function lets the user download a
compressed file of all the images.
processing work. After these changes, the CPU usage
became lower than the constant 100% from before and
the program doesn’t freeze anymore.
Next, the hardware needed for the system were
canvassed then procured. The program was imported
to Raspberry Pi and the actual testing begun. Tests for
the accuracy of the system in terms of ability to detect
multiple entities, low –light conditions, speed limit,
Figure 12. Accounts SMS notification delay, and response time from
The webpage can only be accessed by the admin capture point to the web page notification were
account and the registered users’ credentials shown in conducted for effectiveness and accuracy. The
Figure 12. The admin account is used to carry out researchers also had problems with the functionality of
tasks that require special permissions like adding and the SMS notification function. The GSM module have
deleting users that will be able to access the contents power problems and burst messages. The 5V pin
of the website. When a registered account is logged in, should not be connected by the time Raspberry pi boot
the adding and deleting of accounts feature is disabled up. 5V will be reconnected once Raspberry pi has
and only accessible if the main admin is logged on. booted up. With regard to the burst messages, the
Only the admin account can add and delete user main problem was with the detection of the frames.
accounts. In creating an account, a name, an email, When the system detects a violation, it sends multiple
and a password are needed. After all the fields are messages if the same violator is detected which is not
filled, there is a save button to finalize the user supposed to happen. The problem was solved by
account creation. When deleting an account, the admin making the system detect a violator once and if it is
can browse through the “View All” button and look still in the field of view of the camera, the GSM
for the account you want to delete. Click the trashcan module will not send any extra messages because it is
icon and click “Yes!” on the pop-up to delete the considered as the same motorcycle. The webpage is
selected user. developed after the GSM was fully developed. The
In the product development process, the structuring of frontend side of the development of the website
the project was initiated by conceptualizing the device includes HTML (Hypertext Markup Language), CSS
and the system. After the design was finalized, the (Cascading Style Sheets) and JavaScript. The basic
software development was initialized. The researchers structure of the website like the head and the title was
collected datasets and then trained a model with 2000 developed using HTML. The formatting, layout and
negative images and 1000 positive images. Next, the the overall presentation of the website was managed
model created was used for the development of the using CSS. The backend side of the development of
program for the detection of helmets of motorcycle the website includes PHP and the database
riders. After finalizing the design and training of the development is done with the use of MySQL for the
model, the program for the detection of motorcycle storage of inputs. The researchers then added a login
helmet is completed, the researchers proceeded in feature where only the main admin have the control to
testing it virtually. The code was tested using a laptop add and remove user accounts for security measures.
and a pre-made video. The researchers had a problem After the total assembly of the software and hardware
with the program consuming 100 percent of CPU of the system, multiple tests and evaluations were
usage unnecessarily on the first try. The process was conducted.
monitored and the Raspberry Pi is having difficulties
in running the program because of algorithm
problems. The researchers optimize the program's
flow to avoid extra unnecessary steps. After
optimizing the algorithm of the program, the
researchers were able fix this problem and
successfully reduced the processing power need to run
the system. The researchers made changes like:
isolating the image from the feed then process it as a
cropped image; processing first if the vehicle is indeed
a motorcycle and using multithreading to be able to
distribute the tasks and give the Raspberry Pi easier Figure 13. Multiple Entities Detection
The first part of the multiple detection test is to Table 8. Low-light Conditions Accuracy Test
determine if the device can detect multiple entities at
once. A series of tests was conducted to assess the
ability of the system to detect multiple riders in a
single frame. The researchers initially determined the
maximum number of motorcycles that can be detected
at the same time with consecutive tests by
incrementing the number of entities visible in the same
frame by one.
Table 6. Multiple Entities Detection Test

The proponents conducted the second part of the


multiple detection test, to identify the accuracy of the Figure 14. Correct detection at 5 kph and Blurred
detection with multiple entities. Image at 15 kph
As shown in Table 7, the maximum number of Figure 14 illustrates that the camera can detect a
motorcycles that can be detected at the same time is 3 moving motorcycle with a very low speed (5 kph) and
motorcycles. If the number of motorcycles passing a motorcycle rider moving at 15 kph and faster was
through the camera at the same time is more than 3, visible but it was not a clear vision for detection. The
the system will have difficulty detecting all of the proponents also determined that the system cannot
entities. Table 6 and 7 illustrate the results of the detect a motorcycle rider on a high beam because the
capability of the system in detecting multiple light obstructs the vision of the camera. In conclusion
motorcycle riders. The device can detect at most 3 in these tests, the system can detect violations in low-
motorcycle riders in the same frame. It can also detect light conditions with motorcycles running at a very
violations in different scenarios of three motorcycle low speed. The image of the motorcycle rider is
riders passing through the camera at the same time. blurred when the speed of the motorcycle is 15kph and
Table 7. Multiple Detection Test higher.
Table 9. Speed Limit Test Results

The proponents conducted trials to determine the


maximum speed of a motorcycle the device can detect.
The device was tested through the process of a
The researchers organized tests to determine if the
motorcycle passing through the camera at various
system is capable of operating at night by operating at
speed. The device will have less detection rate if the
low-light conditions. The detection of the system was
motorcycle rider that passes through the camera
tested at different speeds incrementing after each test.
exceeds 20kph.
Table 8 shows the results of the accuracy-test in low-
Table 10. Accuracy Test
light conditions. In comparison, the detection rate in
daylight is significantly higher. Motorcycle riders
running at the speed of 10 kph or more will have a
blurred image interpretation by the camera.

The proponents conducted accuracy tests by the


installation of the device in three different barangays:
The accuracy of these tests was calculated using this
formula:

Table 10 shows the different correct detection results


of the system in each location. Using the formula, the
percentage of accuracy on Barangay Saguin, Barangay
San Roque Dau 1st, and Barangay Ascomo was 74%,
78.29%, and 76% respectively. The proponents
determined that the current accuracy of helmet
detection is 76.09 percent according to the three
accuracy results in different locations.
Table 11. Web Notification Delay Tests

Following the development phase and system testing,


the proponents conducted the system evaluation and
the results are discussed as follows:
Table 13. Respondents Assessment of the Ease of
Operation of the Camera Installation of the
Product

Table 13 represents the assessment of the respondents


with regard to the ease of operation of the Camera
Installation of the Motorcycle Rider Helmet Detection
Table 11 shows the results of the delay of the System. The “camera is easy to focus and aim at
notification through the web server. It was calculated desired area” got the highest weighted mean on
as the time from the motorcycle rider entering the Camera Installation, evaluated as “Excellent” with a
frame of the camera until the time the notification is weighted mean of 3.7. This implies that the camera
shown on the server. The computed average web can be easily set up in order to get the optimal angle
notification delay is 5.4 seconds. for detection. The “The lock of the camera is easy to
As shown in Table 12, the delay of text messages is utilize” and the “The camera is easy to install to
recorded in different instances. There are 20 tests in desired location” got the lowest weighted mean,
total and the time is calculated from the time that the evaluated as “Good” with a weighted mean of 2.8.
motorcycle rider enters the frame until the time that Stronger brackets like metal ones can be used for a
the text message is received by the desired recipient. stronger grip and for an easier lock. The assessed
The computed average delay of the text message quality of the ease of operation of Camera Installation
notification is 12.18 seconds. is “Good” as it obtained the grand mean of 3.1 and it
Table 12. SMS notification delay is within the range of 2.51 – 3.50 based on the Likert
Scale. This suggests that the overall camera
installation is working effectively but it can still be
improved with hardware changes.
Table 14 represents the assessment of the respondents
with regard to the ease of operation of the Device
Controller of the Motorcycle Rider Helmet Detection
System. The “The microcontroller is easy to install to
desired location” got the highest weighted mean on
Device Controller, evaluated as “Good” with a
weighted mean of 3.2. This indicates that the relatively low speed but it is upgradable with higher
Raspberry Pi is easy to install in a specific location specs of a microcontroller.
like a place near a checkpoint operation. The “The
microcontroller runs the program/system fast” got the Table 15. Respondents Assessment of the Process
lowest weighted mean, evaluated as “Good” with a of Notification Receiving of the Product
weighted mean of 2.1. This implies that the
microcontroller does not have the capability of
running the program fast and better hardware is
recommended. The assessed quality of the ease of
operation of the Device Controller is “Good” as it
obtained the grand mean of 2.8 and it is within the Table 16 represents the assessment of the respondents
range of 2.51 – 3.50 based on the Likert Scale. This with regard to the Effectiveness and Accuracy of the
suggests that the installation and the operation of the System in Terms of its Ability to Detect Multiple
device controller are good but the hardware is still Entities, Low – light Conditions, and Notifications on
upgradable. the Motorcycle Rider Helmet Detection System. The
Table 14. Respondents Assessment of the Ease of “system sends notifications to the user after capturing
Operation of the Device Controller of the Product images of violators” got the highest weighted mean,
evaluated as “Excellent” with a weighted mean of 3.6.
This also implies that the system sends notifications
with relative effectiveness and it is working in line
with the objectives of the study. The “result of
multiple detected entities are accurate” had the lowest
mean of 2.6 and this suggests that the system is having
difficulties in detecting multiple entities but it can still
detect three motorcycles in the same frame. Overall,
Table 15 represents the assessment of the respondents the respondents assessed the Effectiveness and
with regard to the process of Notification Receiving of Accuracy of the System in Terms of its Ability to
the Motorcycle Rider Helmet Detection System. The Detect Multiple Entities, Low – light Conditions, and
“user was able to receive notifications” got the highest Notifications as “Good” for it obtained the grand mean
weighted mean on Notification Receiving, evaluated of 3.09, and based on the Likert Scale, it is within the
as “Excellent” with a weighted mean of 3.7. This range of 2.51 – 3.50. The system sends notifications
suggests that the notification of the system is working with effectiveness and the images sent were not
effectively and the users are able to receive the blurred and are labeled with date and time.
notifications in every detection. The “The phone of the
user was able to get access to the webpage” got the Table 16. Respondents Assessment of the
lowest weight mean, evaluated as “Good” with a Effectiveness and Accuracy of the System
weighted mean of 3.1. This implies that the links to
the images are working properly but an external factor
like the absence of internet connection can affect the
viewing of images saved on the webpage. The
assessed quality of the process of Notification
Receiving is “Good” as it obtained a grand mean of
3.4 and it is within the range of 2.51 – 3.50 based on
the Likert Scale. This indicates that the notification of
the system is working well and effectively meet its
objectives. Overall, the respondents assessed the Ease
of Operation of the Product as “Good” for it obtained
the overall grand mean of 3.1, and based on the Likert
Scale, it is within the range of 2.51 – 3.50. This
suggests that the device is user-friendly with a simple
point of operation and the system sends notifications Table 17 shows the assessment results of the
capably. The Raspberry Pi runs the program with professionals who participated in assessing the
Motorcycle Helmet Rider Helmet Detection Using
Image Processing based on ISO/IEC 25010 evaluation that persons or other products or systems have the
questionnaire. degree of data access appropriate to their types and
levels of authorization. The weighted mean of this
Table 17. Experts Assessment for Motorcycle Rider characteristic is 4. The professional respondents are
Helmet Detection Using Image Processing Satisfied with the security of the data stored and the
degree to which the system is only accessible by the
created user using the admin account. The weighted
average of all the characteristics is 4.18 which implies
that the professional respondents are Satisfied with the
overall functionality of the system.
Through a variety of specific tests and evaluations, the
findings of the study were that the system can provide
assistance to the authorities as the results show that the
overall function of the helmet detection system is
Functional Suitability represents the degree to which effective and can serve its intended purpose of
a product or system provides functions that meet providing aid in capturing motorcycle riders who do
stated and implied needs when used under specified not wear a helmet. The results, through evaluation,
conditions. The weighted mean on this characteristic is suggest that the system is functional and is working
4.5. The professional respondents are Satisfied with based on the objectives yet it is upgradable with
how the system detects motorcycle riders accurately hardware and software advancements. Moreover, the
and how the notification provides proper images to the findings on the evaluation of professional respondents
user. Performance Efficiency represents the on the system show satisfaction with the overall
performance relative to the number of resources used function of the system with recommendations in
under stated conditions. The weighted mean of this improving the low-light detection capability, detection
characteristic is 4.11. The professional respondents are rate optimization, and webpage security enhancement.
Satisfied with the detection rate of the device, the The system can only detect the most effective within
capability to detect multiple motorcycle at the same the speed limit of 20kph and the detection rate of the
time and the response time of the system in terms of system is inversely proportional to the speed of the
sending notifications. Usability is the degree to which motorcycle due to processor limitations. In accordance
a product or system can be used by specified users to with the hardware restraints, the device can only
achieve specified goals with effectiveness, efficiency detect a maximum of three motorcycles with optimal
and satisfaction in a specified context of use. The detection. Detection in low-light conditions was less
weighted mean of this characteristic is 4.67. The effective compared with the detection rate in optimal
professional respondents are Very Satisfied with the conditions. The proponents conducted a comparison
ease of use of both the system and the webpage and testing between raspberry pi and a laptop, the
with the ease of navigation of the webpage. efficiency of the detection rate of the laptop is
Reliability is the degree to which a system, product or significantly more effective. The calculated average
component performs specified functions under delay of the notifications of the webpage and the text
specified conditions for a specified period of time. The messages are 5.4 seconds and 12.18 seconds,
weighted mean of this characteristic is 3.56. The respectively. The average accuracy of the device from
professional respondents are Satisfied with the tests with a total of 1050 motorcycles is 76.09%
capability of the device in low-light conditions, the
accuracy of the device on detecting multiple IV. Conclusion
motorcycle riders and with the data storage The researchers concluded that Motorcycle Rider
management of the system. Portability is the degree Helmet Detection Using Image Processing achieved
of effectiveness and efficiency with which a system, its objectives in terms of functionality. The officers
product or component can be transferred from one who are involved in checkpoint operations have
hardware, software or other operational or usage assessed that the system is easy to use, efficient, and
environment to another. The weighted average mean able to help keep evidence of violators. The system
of this characteristic is 4.33. The professional was able to detect no-helmet motorcycle riders and
respondents are Satisfied with the efficiency of the sends a notification through a website and text
device installation. Security is the degree to which a message. In accordance with the tests results and
product or system protects information and data so evaluations, the end-user respondents assessed the
system as “Good” and the experts were “Satisfied”
with the overall function of the system with [1] Liu et al. (2008). Helmets for preventing injury in motorcycle
riders. Research Gate.
recommended improvements in terms of optimization https://www.researchgate.net/publication/8599262_Helmets_for_pr
of the detection rate of the raspberry pi, improvement eventing_injury_in_motorcycle_riders
on low-light detection capability, and webpage [2] LawPhil. (2009). REPUBLIC ACT NO. 10054. The LawPhil
Project.
security.
https://lawphil.net/statutes/repacts/ra2010/ra_10054_2010.html
[3] Pabaharan. T.2020. Studies on application of image
V. Recommendation processing in various fields: An overview.
The product of the study was able to meet the https://sisu.ut.ee/imageprocessing/book/1
requirements in terms of its purpose and features. The [4] Waranusast et al. (2013). Machine vision techniques for
motorcycle safety helmet detection. IEEE.
system can still be improved by upgrading hardware https://ieeexplore.ieee.org/abstract/document/6726989/authors#aut
components. A more powerful microprocessor can hors
enhance the speed of the system. Subsequently, a [5] Siebert, F., & Lin, H.2019. Detecting motorcycle helmet use
with deep learning.
better camera will also improve the resolution of the https://www.researchgate.net/publication/340941340_Detecting_m
captured photo, so other details can be clearer and the otorcycle_helmet_use_with_deep_learning
captured image will not be blurred. The webpage can [6] WHO PH. (2018). WHO PH: Over 90% of Motorcycle
Deaths Didn’t Wear Helmets. Carmudi.
still be improved by further development like adding a https://www.carmudi.com.ph/journal/philippines-90-motorcycle-
filter function where the user can filter the date and deaths-didnt-wear-helmets/
time of the saved images for faster navigation. The [7] McCombes, Shona,2020. Descriptive Research Design |
security of the website can still be upgraded to prevent Definition, Methods and Examples. Transportation Based on Public
Perception in Developing Countries. Science
internet breaches and avoid loss or theft of data. https://www.hindawi.com/journals/jat/2020/9084245/
[8] IStock.2022. Man motorcycle. Retrieved from:
VI. Acknowledgements https://www.istockphoto.com/en/search/2/image?
mediatype=photography&phrase=man+motorcycle&utm_source=pi
This section aims to give gratitude to all of the
xabay&utm_medium=affiliate&utm_campaign=ADP_photo_spons
individuals that helped financially, emotionally and ored&utm_content=http%3A%2F%2Fpixabay.com%2Fphotos
mentally and spiritually in the course of the study. %2Fman-motorcycle-rider-helmet-
Asil Kastle Dela Cruz, MIT, PCpE – for being the 7028819%2F&utm_term=man+motorcycle\
[9] Pixabay.2022. Motorcycle rider with helmet. Retrieved from:
researchers’ thesis and academic adviser, for guiding
https://pixabay.com/images/search/motorcycle%20rider%20with
them, giving them feedback on what improve and %20helmet/
supporting them in completing the project. [10] Shutterstock.2022. Motorcycle rider with helmet royalty-free
Brgy. Captain Richard A. Manalili – for giving the images. Retrieved from:
https://www.shutterstock.com/search/motorcycle-riderwithhelmet?
researchers the permission to collect data inside
kw=stock
barangay Saguin, City of San Fernando, Pampanga. %20image&c3apidt=p69380673812&gclid=CjwKCAjwk_WVBhB
Larry Page & Sergey Brin – for developing google ZEiwAUHQCmRCyrNLQg7N5JEceDlRFblU6GjMkP32m0LpFC8
that provided the researchers guides and information ANQoIUowUrhY9hoCNdoQAvD_BwE&gclsrc=aw.ds
[11] Unsplash. (2022). Motorcycle rider with helmet. Retrieved
they needed in completing their frameworks, tasks and from: https://unsplash.com/s/photos/motorcycle-rider-with-helmet
module preparation.
Traffic authorities – for giving the researchers answers
that will help their research paper to make more
validation.
Maribel Valerio – for providing the researchers a place
to stay when they need an overnight stay to discuss the
study.
Other family and friends – for giving the researchers
the emotional, moral and emotional support they
needed in completing the research paper.
Almighty God - For his ever-lasting love and grace to
give us even things that we don’t even deserve, for
giving us the inner spiritual strength to stay on the
path righteousness, for giving us wisdom and
knowledge to accomplish everything at hand.

VII. References

You might also like