I

ABSTRACT

As the rate of urbanization continues to rise, the problem of traffic management will become more important

and will call for creative solutions. This research provides a revolutionary approach to the automation of

traffic signals by capitalizing on the combination of cutting-edge computer vision methods and technology

that is enabled by the Internet of Things (IoT). Two specialized nodes, each of which is outfitted with the

powerful YOLOv3 object detection model, have been painstakingly created in order to distinguish and

measure the volume of vehicle traffic at signal intersections with an accuracy that is unmatched. The

information that has been identified is then sent in a smooth manner to a central server that is housed on a

Raspberry Pi, which results in the construction of a network that is nimble and responsive. This server acts

as the command center and is responsible for coordinating the cycle of the traffic signals depending on the

current real-time circumstances. The system optimizes traffic flow dynamically, hence reducing congestion

and increasing overall urban mobility, by intelligently distributing green light time in direct connection to

the observed vehicle density. This helps the system mitigate congestion and improves overall urban

mobility. Not only does this comprehensive combination of cutting-edge computer vision with the Internet

of Things connection exemplify a jump in technical innovation, but it also represents a paradigm change in

the way that intelligent and responsive traffic management systems operate. The ramifications of this

system are significant, and it has the potential to bring about a revolution in the infrastructure of urban

transportation and to establish new standards of urban sustainability, efficiency, and safety.

II
 ACKNOWLEDGEMENT

To begin, we would want to express our gratitude to Allah, who has been there for us from the moment we

were born to the present day, assisting us in each and every stage of our lives. We are grateful to everyone

who had a part in assisting us in reaching our objective of having a successful project during the course of

the year. Without them, we were unable to complete any of the project's duties, no matter how trivial or

important they were.

We would like to express our gratitude to our parents for all of the support and encouragement they have

given us. All of us want to express our gratitude to our cherished parents via this effort. Without their

assistance, we would not have been able to finish the project, since they gave us all of the resources, we

could possibly need without any conditions attached.

We are eternally grateful to our internal advisor __________________ for all the advice, guidance, and

valuable support he gave us throughout our final year project. His guidance and encouragement will help

us in our professional careers.

S.No. Student Name Roll number

III
Table of Contents

ABSTRACT ................................................................................................................................................. II
ACKNOWLEDGEMENT ........................................................................................................................... III
1. INTRODUCTION ................................................................................................................................. 1
2. BACKGROUND ................................................................................................................................... 3
3. HARDWARE USED ............................................................................................................................. 4
3.1. Raspberry Pi 4................................................................................................................................ 4
3.1.1. Key Specifications: ................................................................................................................ 4
3.1.2. Operating Systems: ................................................................................................................ 5
3.2. CAMERA ...................................................................................................................................... 9
3.3. LEDs ............................................................................................................................................ 10
3.4. Rasbian as an Operating System.................................................................................................. 11
4. Hardware and Software Implementation ............................................................................................. 13
4.1. YoloV3 ......................................................................................................................................... 14
4.2. Internet of Things (IoT) ............................................................................................................... 17
5. Results.................................................................................................................................................. 19
6. Code ..................................................................................................................................................... 21
7. Conclusion ........................................................................................................................................... 23
References.................................................................................................................................................... 25

IV
 1. INTRODUCTION

The extraordinary increase in automobile traffic that has been ushered in as a result of urbanization has

presented urban planners and traffic management agencies all over the globe with significant issues [1]. It

is becoming more necessary to have traffic management systems that are both efficient and able to react to

changing conditions as cities continue to develop and their populations increase. Conventional traffic signal

systems, which are often regulated by pre-defined time schedules, find it difficult to adapt to the dynamic

nature of traffic patterns. As a consequence, this results in chronic congestion, longer commuting times,

and an increased environmental effect. This project aims to solve this significant problem by using an

innovative combination of cutting-edge computer vision methods and Internet of Things (IoT) technology

[2-4]. The eventual result of this attempt will be an automated traffic light system that is capable of

providing a real-time response to changes in the flow of traffic [5].

The use of the YOLOv3 (You Only Look Once) object detection model, which is recognized for its

extraordinary speed and accuracy in recognizing and categorizing objects in photos, is essential to the

success of this attempt. YOLOv3 is used in this scenario for the purpose of precisely identifying and

quantifying the number of cars that are present at signalized junctions. Because it makes use of specialized

nodes that are outfitted with this cutting-edge detection model, the system is able to attain a degree of

accuracy that has not been seen before in traditional methods of traffic monitoring. These nodes function

as the system's eyes, gathering and processing visual input in real-time. As a result, they create the basis

upon which future choices about traffic management are founded [6].

This automated traffic signal system utilizes a seamless integration of Internet of Things (IoT) technology,

which acts as the nervous system. This integration enables quick and reliable communication between the

vehicular detecting nodes and a centralized server. This server, which is housed on a Raspberry Pi, operates

as the center for decision making by analyzing the incoming data streams and coordinating the signal cycle

in accordance with those interpretations. Even in situations with a high data flow, the system is able to

1
reliably and effectively share information because it makes use of common Internet of Things

communication protocols [7]. One example of such a protocol is MQTT, which stands for Message Queuing

Telemetry Transport. This infrastructure, which is enabled by the Internet of Things, not only makes it easier

for the system to respond in real-time but also paves the way for possible scalability and flexibility to a

wide variety of urban settings [8].

In this fully automated traffic light system, the Raspberry Pi, which is famous for its adaptability and

processing capability even in limited settings, plays a crucial part. In addition to playing the role of the

primary server, the Raspberry Pi also plays the role of a safe and dependable platform that is used to house

the necessary software and algorithms [9]. This device, which is both small and powerful, plays an

important role in the processing of the incoming data streams from the vehicular detecting nodes, the

execution of the required calculations, and the generation of the relevant control signals to govern traffic

flow. Because of its low power consumption and cost-effectiveness, it is a great choice for large-scale

deployment, and it holds the potential to revolutionize traffic control systems on a scale that is applicable

to a whole city [10].

In conclusion, this project proposes a comprehensive and synergistic approach to the automation of traffic

signals. It addresses the modern issues of urban traffic management by integrating cutting-edge computer

vision and Internet of Things technologies. This cutting-edge system will usher in a new age of intelligent

and responsive traffic control thanks to the convergence of vehicle detection nodes based on YOLOv3, a

central server based on Raspberry Pi, and IoT communication protocols [11]. This convergence will

constitute the backbone of this unique system. The succeeding parts are going to go into the technical

subtleties of each component, illuminating the design considerations, implementation details, and

performance evaluations, which will finally culminate in a full assessment of the system's efficacy in real-

world urban areas [12].

2
 2. BACKGROUND

The need for traffic management systems that are both effective and flexible is becoming more urgent as

urban areas continue to expand. The traditional traffic signal systems, which are regulated by set time

schedules, sometimes fail to accommodate the dynamic nature of traffic patterns. As a consequence, this

results in chronic congestion, higher travel times, and an increased environmental impact. A paradigm shift

toward creative solutions that are able to respond in real-time to changing traffic circumstances is required

as a result of this need. The advent of computer vision technology, defined by breakthroughs in object

identification and image processing, presents a potential route for transforming traffic management [13-

14]. This might be accomplished in a number of ways. Among these breakthroughs, the YOLOv3 (You

Only Look Once) object detection model has garnered international notice for its amazing speed and

accuracy in distinguishing and categorizing objects inside photos. The model's name comes from the phrase

"You only look once." This project aims to establish the foundation of an intelligent traffic management

system by using the power of YOLOv3 in order to improve the accuracy and effectiveness of vehicle

identification at signalized junctions. This will allow the system to function as the central pillar [15].

A new age of connectedness and data sharing has been ushered in as a result of the widespread adoption of

Internet of Things (IoT) technology, which has occurred in parallel with these technological breakthroughs.

The Internet of Things is a very important enabler for the traffic control system's many components to

communicate with one another in a seamless manner. The system provides the degree of resilience and

dependability essential for decision-making in real time via the adoption of established communication

protocols such as MQTT. This is important for real-time decision-making. This use of Internet of Things

technology not only contributes to the responsiveness of the system, but it also sets the groundwork for the

system's future scalability and adaptation to a wide variety of urban settings [16].

This integrated system is improved further with the addition of the Raspberry Pi, which is a computer

platform that is both flexible and affordable. In addition to its function as the primary server, the Raspberry

3
Pi plays a critical part in the processing and interpretation of the data streams as they come in. Because of

its computing power in contexts with limited resources, it is a great choice for large-scale deployment, and

it has the potential to revolutionize traffic management systems on a scale that is applicable to a whole city.

This project aims to overcome the constraints of traditional traffic management systems by integrating these

technological breakthroughs in a seamless manner [17]. As a result, it provides a dynamic and intelligent

approach to the automation of traffic signals. The subsequent sections will delve into the technical

complexities of each component, providing detailed insights into the design considerations, implementation

specifics, and performance evaluations, which will ultimately culminate in a comprehensive assessment of

the effectiveness of the system in actual world urban environments.

3. HARDWARE USED

Following are the hardware components used for the project development,

• Raspberry Pi 4

• Camera

• LEDs

• Rasbian as an Operating System

3.1. Raspberry Pi 4

The Raspberry Pi 4, released in June 2019, is the latest iteration of the highly popular single-board computer

series developed by the Raspberry Pi Foundation. Building on the success of its predecessors, the Raspberry

Pi 4 introduces significant advancements in performance, connectivity, and versatility, making it a versatile

and powerful platform for a wide range of applications.

3.1.1. Key Specifications:

4
 ▪ Processor: The Raspberry Pi 4 is equipped with a quad-core Broadcom BCM2711 processor based

on the ARM Cortex-A72 architecture. Running at 1.5 GHz, this processor provides a substantial

increase in processing power compared to earlier models.

▪ RAM: It is available in various configurations, with options for 2GB, 4GB, or 8GB of LPDDR4

SDRAM. This expanded memory capacity enables smoother multitasking and more demanding

applications.

▪ Graphics: The Pi 4 features a VideoCore VI GPU, offering improved graphics performance for

tasks ranging from high-definition video playback to 3D rendering.

▪ Connectivity: It boasts comprehensive connectivity options, including two USB 3.0 ports, two USB

2.0 ports, Gigabit Ethernet, dual-band Wi-Fi (802.11ac), and Bluetooth 5.0. These enhancements

greatly expand the range of peripherals and devices that can be connected to the Raspberry Pi 4.

▪ Video Output: The Pi 4 supports dual 4K display output via two micro HDMI ports. This capability

is particularly valuable for applications that require high-resolution visual output.

▪ Storage: It employs a microSD card slot for primary storage, offering ample space for the operating

system and applications. Additionally, USB and network-based storage options can be utilized for

expanded storage capacity.

▪ Power: The Raspberry Pi 4 requires a 5V USB-C power supply. The power requirements are

slightly higher compared to previous models due to the increased processing capabilities.

3.1.2. Operating Systems:

The Raspberry Pi 4 is compatible with a wide range of operating systems, including Raspbian (now known

as Raspberry Pi OS), Ubuntu, and various Linux distributions. It can also run specialized operating systems

for specific applications, such as media centers, retro gaming consoles, and robotics platforms.

• Use Cases:

5
The enhanced processing power and expanded memory of the Raspberry Pi 4 make it suitable for a diverse

array of applications. These include but are not limited to:

▪ Desktop Computing: With its improved performance, the Pi 4 can serve as a basic desktop computer

for web browsing, word processing, and other office tasks.

▪ Media Center: It can function as a capable media center, capable of streaming high-definition video

and supporting a variety of multimedia formats.

▪ IoT and Embedded Systems: Its small form factor and GPIO pins make it an ideal platform for IoT

projects, sensor interfacing, and home automation.

▪ Educational Tools: The Raspberry Pi 4 continues the tradition of providing an accessible and

affordable platform for learning programming and computer science.

Fig.1. Raspberry Pi 4 Model

6
Fig. 2. Raspberry Pi Pin Configuration

7
• Pin Numbering:

• The GPIO pins on the Raspberry Pi 4 are labeled with both physical pin numbers and

GPIO numbers. It's important to note that there are two different numbering schemes:

• Physical Pin Numbering: This refers to the actual pin on the Raspberry Pi's

GPIO header.

• BCM GPIO Numbering: This numbering scheme is based on the Broadcom

SOC channel numbers. These are the numbers used in programming to interact

with the GPIO pins.

• Power Pins:

• 3.3V (Pin 1): This pin provides a 3.3V power supply, which can be used to power low-

voltage components.

• 5V (Pin 2): This pin provides a 5V power supply, which can be used to power higher-

voltage components.

• Ground Pins:

• GND (Pins 6, 9, 14, 20, 25, 30, 34, 39): These pins are ground connections, providing a

reference voltage for circuits.

• GPIO Pins:

• There are a total of 40 GPIO pins on the Raspberry Pi 4, labeled as GPIO0 through

GPIO39 in BCM numbering.

• These pins can be used for both input (reading signals) and output (sending signals)

purposes.

• I2C Pins:

• SDA (Pin 3): This is the data line for I2C communication.

• SCL (Pin 5): This is the clock line for I2C communication.

• SPI Pins:

• MOSI (Pin 19): Master Out Slave In for SPI communication.

8
 • MISO (Pin 21): Master In Slave Out for SPI communication.

• SCLK (Pin 23): Serial Clock for SPI communication.

• CE0 (Pin 24): Chip Enable 0 for SPI communication.

• CE1 (Pin 26): Chip Enable 1 for SPI communication.

• UART Pins:

• TXD (Pin 8): Transmit Data for UART communication.

• RXD (Pin 10): Receive Data for UART communication.

• PWM Pins:

• PWM0 (Pin 12): Pulse Width Modulation (PWM) output.

• PWM1 (Pin 32): Pulse Width Modulation (PWM) output.

• Other Pins:

• ID_SD (Pin 27): I2C ID EEPROM (used to identify the board and load the correct

drivers).

• ID_SC (Pin 28): I2C ID EEPROM (used to identify the board and load the correct

drivers).

• Reserved Pins:

• Some pins are reserved for specific functions (e.g., EEPROM, HAT ID) and should not

be used for general-purpose applications.

3.2. CAMERA

The camera component is an integral aspect of this project, facilitating real-time visual data acquisition and

processing. The camera employed is a high-resolution module designed to interface seamlessly with the

Raspberry Pi 4, leveraging its capabilities for a diverse range of applications. It is used to capture the live

video stream of traffic with can be used for a variety of purposes.

9
 Fig. 3. Camera

3.3. LEDs

Light Emitting Diodes, frequently referred to simply as LEDs, are semiconductor devices that, when an

electric current is sent through them, produce light. They have a positive lead (the anode) and a negative

lead (the cathode), and in order for them to operate, they must be linked in the appropriate direction. LEDs

are available in a wide range of hues, including red, green, blue, and white, with each color requiring a

unique combination of semiconductor elements to create. They are often used in the electronic industry for

the purpose of providing a visible signal, such as in the form of power indicators or status lights. In

comparison to conventional incandescent bulbs, light-emitting diodes, or LEDs, have a longer lifetime,

require less power, and are more resilient.

10
 Fig. 4. LEDs

3.4. Rasbian as an Operating System

▪ Purpose-Built for Raspberry Pi:

Raspbian is developed and maintained by the Raspberry Pi Foundation, with the primary goal of providing

an operating system that is tailored to the hardware and capabilities of Raspberry Pi devices.

▪ Debian-Based Linux Distribution:

Raspbian is derived from Debian, a popular and widely used Linux distribution known for its stability,

extensive software repositories, and strong community support.

▪ Optimized Performance:

Raspbian is optimized to run efficiently on Raspberry Pi hardware, making the most of its processing power

and memory capabilities.

▪ Desktop Environment:

Raspbian includes a lightweight desktop environment, typically based on the LXDE or LXQt desktop

environments. This provides users with a graphical user interface (GUI) for easy navigation and interaction

with the system.

▪ Package Management:

11
Raspbian uses the Debian package management system, which allows users to easily install, update, and

manage software packages using package managers like apt.

▪ Vast Software Repository:

Raspbian provides access to a vast repository of pre-compiled software packages, which cover a wide range

of applications and utilities. This makes it easy for users to install additional software on their Raspberry

Pi.

▪ GPIO Access:

Raspbian includes tools and libraries that allow users to interact with the GPIO pins of the Raspberry Pi,

enabling the development of hardware projects and IoT applications.

▪ Education and Learning:

Raspbian is well-suited for educational purposes and serves as an excellent platform for learning

programming, electronics, and computer science. It comes pre-installed with educational software and

tools.

▪ Regular Updates and Support:

Raspbian receives regular updates and security patches, ensuring that users have access to the latest features

and improvements. The Raspberry Pi community and forums provide additional support and resources.

▪ Customizations and Configurations:

Users have the flexibility to customize and configure Raspbian to suit their specific needs, allowing for a

tailored computing experience.

▪ Compatibility:

Raspbian is designed to be compatible with a wide range of Raspberry Pi models, from the earliest versions

to the latest releases.

12
 ▪ Variants of Raspbian:

There are different variants of Raspbian available, including the full desktop version (with GUI), a lite

version (without GUI, suitable for headless setups), and specialized variants like Raspbian for Robots.

4. Hardware and Software Implementation

The hardware solution for this project is centered on the incorporation of two specialized vehicle-detecting

nodes, each of which is powered by a Raspberry Pi 4 and comes equipped with an onboard camera [18].

The powerful YOLOv3 object identification model is used by these nodes, which play an important part in

the process of obtaining real-time visual data of the road. This allows the nodes to precisely determine and

quantify the number of cars that are present at signalized junctions [19-23]. The YOLOv3 model, which is

recognized for its excellent speed and accuracy, is used to analyze the acquired photographs in order to

provide exact information on the density of the traffic. These vehicle-detecting nodes are built with Internet

of Things characteristics, which enables them to interact fluidly with the central server that is located on a

separate Raspberry Pi 4 device. The MQTT protocol, which is a lightweight and efficient messaging system

that is perfect for Internet of Things applications, is what enables communication to take place. Because of

this contact, real-time updates on the traffic situation are guaranteed, which is an essential component of

the project's operation. The server software, which is stored on the Raspberry Pi, is in charge of receiving,

analyzing, and processing the data that is sent by the vehicle-detecting nodes. It makes judgments on the

management of the traffic signals based on this data in order to be as informed as possible. The Raspberry

Pi server improves overall urban mobility by optimizing traffic flow. It does this by dynamically modifying

the signal cycle depending on the observed vehicle density. This helps to alleviate congestion and improves

overall urban mobility. In addition, suitable power management systems and secure mounting enclosures

are installed in order to guarantee steady functioning and shield the components from the effects of the

surrounding environment. The integration of these different pieces of hardware in such a way as to be

13
carefully choreographed constitutes the basis of an intelligent traffic management system, which is on the

verge of revolutionizing the infrastructure of urban transportation [24-25].

Fig. 5. Hardware Implementation using Internet of Things (IoT)

4.1. YoloV3

You Only Look Once, more usually abbreviated as YOLO, is an innovative strategy for object recognition

that was developed in the field of computer vision. YOLO is a programming language that was created by

Joseph Redmon and Ali Farhadi. It is notable for both its efficiency and speed, which makes it especially

well-suited for real-time applications. The third version of YOLO, also known as YOLOv3, is an

improvement over previous versions of the software in terms of both its accuracy and its performance [18].

YOLOv3 fundamentally changes the way object detection is approached by recasting it as a problem

involving regression. Traditional approaches for detecting objects in images include dividing a picture into

grids, applying sophisticated region recommendations, and creating classification networks. On the other

hand, YOLO approaches object detection as a unified and comprehensive whole. It eliminates the need for

numerous calculations by dividing the picture into a grid and predicting bounding boxes and class

14
probabilities immediately. Because of this, the procedure of detection is made extraordinarily effective and

simplified [17].

There are many distinguishing characteristics that define the YOLOv3 architecture. The first thing that it

does is use a deep convolutional neural network as its backbone. This network is often a variation of

Darknet. The task of extracting characteristics from the picture that is sent into this network is its

responsibility. YOLOv3 makes use of a number of convolutional layers, which gives it the ability to learn

hierarchical representations of objects at varying sizes. This is an essential capability for effective detection.

In addition, the YOLOv3 model makes use of various detection scales, a technique known as "multi-scale

detection," to increase the model's capacity to recognize objects of varied sizes. Within the network, it

employs detection at three distinct scales, so guaranteeing that tiny, medium, and big items are all reliably

detected [26].

In addition to this, YOLOv3 implements a notion that is referred to as "anchor boxes." The model relies on

these pre-defined bounding boxes, which might be of varying sizes and aspect ratios, to make accurate

predictions about the placements of objects. Because it makes use of anchor boxes, YOLOv3 has the ability

to conform to objects of varying sizes and forms, which contributes to the diversity of the system [19].

YOLOv3 has gained the capability of predicting item classes at different scales, which is another significant

improvement. This indicates that it is able to recognize not just the existence of an item but also the

particular kind or category of that object, which significantly improves its capabilities when dealing with

complicated scenarios.

YOLOv3 is a pioneering object detection model that redefines the paradigm by approaching object

detection as a single regression problem. In a nutshell, this makes YOLOv3 an innovative approach to

object identification. Its cutting-edge design, which includes multi-scale detection, anchor boxes, and class

predictions, makes it possible to identify objects in real-time settings with high accuracy and efficiency.

Because of its ability to strike a good balance between speed and accuracy, YOLOv3 has become an

15
essential component in a wide variety of computer vision applications. One such application is the

identification of automobiles on roadways, which is what this project is focusing on [21].

Fig. 6. YoloV3 Pretrained Classifier

16
4.2. Internet of Things (IoT)

The Internet of Things, often known as IoT and usually abbreviated as "IoT," is a paradigm shift in terms

of how technology and connection are conceptualized and used. It is a network of linked systems, gadgets,

and sensors that interact with one another and exchange data through the internet. This dynamic network

offers a broad variety of applications, ranging from industrial automation and smart homes to healthcare

and urban infrastructure, which significantly alters the manner in which we interact with our surroundings.

The Internet of Things is based, fundamentally, on the seamless integration of various software and

hardware components. The 'things' that make up the Internet of Things are defined as any electronic devices

that have both sensors and communication modules. These sensors are able to collect a wide variety of data,

including temperature and humidity readings, as well as data on motion and light levels. After the data has

been gathered, it is then sent to a centralized hub or server via a variety of communication protocols,

including Wi-Fi, Bluetooth, Zigbee, or cellular network connections [25].

The capability of these devices to function independently, often with just minimum input from a human

operator, is essential to the usefulness of the Internet of Things. They are able to analyze data locally, come

to judgments that are informed based on criteria that have been set, and even connect with other devices

that are part of the network. This degree of independence gives Internet of Things systems the ability to

react quickly to changes in their surroundings, which paves the way for real-time control and automation.

The capability of the Internet of Things to provide remote monitoring and control is one of its defining

characteristics. This skill has a wide range of applications across a variety of business sectors, ranging from

agriculture and healthcare to manufacturing and smart cities. For instance, sensors that are connected to the

internet of things may be used in agriculture to monitor the moisture levels and temperature of soil. This

gives farmers the ability to optimize irrigation and change planting dates. In a similar vein, Internet of

Things-enabled medical devices in the field of healthcare may provide important patient data to healthcare

practitioners in real time, so allowing prompt treatments.

17
When implementing IoT solutions, security and privacy concerns should be given top priority. Because of

the vast number of interconnected devices that are constantly transferring sensitive data, it is vital to

implement stringent security measures to guard against unauthorized access and breaches of data. When it

comes to protecting the integrity of an Internet of Things system, encryption, secure authentication

procedures, and consistent software upgrades are all essential components.

Additionally, one of the unique characteristics of IoT is its scalability. IoT ecosystems may consist of a very

small number of devices all the way up to hundreds or even millions of nodes and devices. The Internet of

Things can accommodate a wide variety of applications because to its scalability, ranging from individual

smart home installations to enormous industrial deployments.

Fig. 7. Internet of Things (IoT)

18
5. Results
YOLOv3, which stands for "You Only Look Once," is a state-of-the-art object detection model that excels

in identifying objects within images or frames of a video. Its core innovation lies in its remarkable speed

and accuracy, making it ideal for real-time applications, including the detection of objects like cars.

The YOLOv3 object detection process begins with the division of the input image into a grid of cells. Each

cell is responsible for predicting objects within its boundaries. Within each cell, YOLOv3 predicts bounding

boxes that encapsulate the objects it detects, along with the confidence score of the prediction.

Crucially, YOLOv3 is capable of identifying multiple objects in a single image, and it can predict the class

(e.g., car, person, bicycle) of each detected object. This is achieved through the use of anchor boxes, which

are predefined bounding boxes of varying sizes and aspect ratios. YOLOv3's network predicts the

coordinates of these anchor boxes relative to each cell, and each anchor box is associated with a specific

class of object.

To understand how YOLOv3 detects cars in an image:

Feature Extraction: YOLOv3 utilizes a deep convolutional neural network (CNN) as its backbone to extract

features from the input image. These features capture various patterns, textures, and shapes that help in

object recognition.

Bounding Box Prediction: For each grid cell, YOLOv3 predicts multiple bounding boxes (typically 3 or 9).

These bounding boxes represent the potential locations of objects within the cell.

Object Confidence: For each predicted bounding box, YOLOv3 assigns a confidence score, indicating how

likely it is that there is an object inside the box. High confidence values indicate a strong belief in the

presence of an object.

19
Class Prediction: YOLOv3 also predicts the class probabilities for each bounding box. In the context of

detecting cars, it assigns a higher probability to the "car" class if a car is present within the bounding box.

Non-Maximum Suppression: After predictions are made for all grid cells, YOLOv3 applies a technique

called non-maximum suppression (NMS) to filter out duplicate and low-confidence detections. This ensures

that only the most confident and accurate detections are retained.

The result is a set of bounding boxes, each associated with a confidence score and a class label. In the

context of your project, YOLOv3 identifies and highlights the cars present in the image, providing their

locations and class labels. This information can then be used for further analysis or action, such as traffic

management based on the number of detected cars. YOLOv3's efficiency and accuracy make it a powerful

tool for object detection in a wide range of applications, including traffic monitoring and control.

Fig. 8. Sample 1 output of YoloV3 pretrained classifier

20
Fig.9. Sample 2 output of YoloV3 pretrained classifier

6. Code

import cv2
import numpy as np
import requests
import random as rn

# Load the YOLO model

net = cv2.dnn.readNet("yolov3.weights", "yolov3.cfg")
classes = []
with open("coco.names", "r") as f:
classes = [line.strip() for line in f.readlines()]
layer_names = net.getLayerNames()
output_layers = [layer_names[i[0] - 1] for i in net.getUnconnectedOutLayers()]

# Open the video capture

cap = cv2.VideoCapture(0) # Change the argument to a video file path if needed

car_count = 0

21
prev_car_count = 0

# Raspberry Pi 4 information
pi4_ip = "192.168.1.104:5000" # Replace with the actual IP address of your Raspberry Pi 4

while True:
# Read the video frame
ret, frame = cap.read()

# Perform object detection

height, width, channels = frame.shape
blob = cv2.dnn.blobFromImage(frame, 0.00392, (416, 416), (0, 0, 0), True, crop=False)
net.setInput(blob)
outs = net.forward(output_layers)

# Process the detected objects

class_ids = []
confidences = []
boxes = []
for out in outs:
for detection in out:
scores = detection[5:]
class_id = np.argmax(scores)
confidence = scores[class_id]
if confidence > 0.5 and class_id == 2: # Class ID 2 corresponds to 'car'
center_x = int(detection[0] * width)
center_y = int(detection[1] * height)
w = int(detection[2] * width)
h = int(detection[3] * height)
x = int(center_x - w / 2)
y = int(center_y - h / 2)
boxes.append([x, y, w, h])
confidences.append(float(confidence))
class_ids.append(class_id)

# Apply non-maximum suppression to remove overlapping bounding boxes

indexes = cv2.dnn.NMSBoxes(boxes, confidences, 0.5, 0.4)

# Draw bounding boxes and labels on the frame

s2_car_count = len(indexes)
#s1_car_count = rn.randint(1,5)

for i in range(len(boxes)):
if i in indexes:
x, y, w, h = boxes[i]
label = f"{classes[class_ids[i]]}: {confidences[i]:.2f}"
cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 2)

22
 cv2.putText(frame, label, (x, y - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)

# Display the car count

if s2_car_count != prev_car_count:
#if s2_car_count >= 2:
print("Number of cars:", s2_car_count)
#requests.get(f"http://{pi4_ip}/S2_on_M")
requests.get(f"http://{pi4_ip}/S2?s2_car_count={s2_car_count}")
#else:
#print("Number of cars:", s2_car_count
#requests.get(f"http://{pi4_ip}/S2_on_L")
#requests.get(f"http://{pi4_ip}/S2_on_L?s2_car_count={s2_car_count}")

prev_car_count = s2_car_count

# Show the resulting frame

cv2.imshow("Car Detection", frame)

# Exit the loop if 'q' is pressed

if cv2.waitKey(1) & 0xFF == ord('q'):
break

# Release the video capture and close the window

cap.release()
cv2.destroyAllWindows()

7. Conclusion
The introduction of an automated traffic signal system that makes use of YOLOv3 object detection and

Internet of Things technology marks a major step forward in the management of traffic in metropolitan

areas. This project incorporates cutting-edge technology in a seamless manner in order to increase the

effectiveness and responsiveness of traffic signals, which will eventually lead to improvements in urban

transportation. The system achieves amazing accuracy in recognizing and measuring automobiles at

signalized crossings because it uses YOLOv3, a cutting-edge object identification model. This allows for

the system to fulfill its goals. The use of this technique, which "looks once" at the complete picture in order

to create predictions, is critical for doing real-time traffic analysis. The resilience and flexibility of YOLOv3

are shown by the system's capability of accurately detecting vehicles regardless of the size or orientation of

23
the targets. In addition to this, the inclusion of IoT makes it possible for the vehicle-detecting nodes and the

central server, which is based on Raspberry Pi, to communicate in an easy and smooth manner. Because of

this communication backbone, the traffic signal system is able to maintain its awareness of the current real-

time traffic circumstances. It has become clear that the MQTT protocol, which was selected due to its ability

to provide lightweight and efficient messaging, is a critical factor in allowing speedy and reliable data

transfer. The Raspberry Pi acts as the system's brain and is responsible for coordinating the operation of

traffic signals depending on the information that is received. The system is able to maximize the flow of

traffic and reduce the amount of congestion by dynamically altering the timing of the signals in response to

variations in the density of the traffic. This flexibility is a key trait, particularly in dynamic metropolitan

contexts where traffic conditions may swiftly shift from one moment to the next. The hardware

implementation of the project, which has been thoroughly described, demonstrates the smooth integration

of many components. These components range from automotive detection nodes that are equipped with

onboard cameras to sophisticated Raspberry Pi servers. The efficiency of the system as a whole is dependent

on the contributions made by each component, which all play an essential part. In addition, aspects like as

power management, enclosures, and secure connections illustrate an all-encompassing strategy for

guaranteeing the system's dependability and longevity. The fact that this traffic management system was

able to pass both its installation and testing phases successfully demonstrates that it has the potential to be

used in a variety of settings. In addition to its immediate use case, this project sets the framework for

additional developments in urban traffic systems. It provides a solution that is scalable and adaptive for

cities that are dealing with expanding transportation difficulties. In conclusion, the incorporation of

YOLOv3 object identification and IoT technology into this automated traffic light system demonstrates the

enormous potential of cutting-edge technologies to transform the way urban traffic management is carried

out. The accomplishments of the project, which include successfully identifying cars and dynamically

managing traffic signals, are evidence that this integrated strategy is effective. These kinds of solutions

represent a potential step toward a smarter and more effective urban transportation system at a time when

cities are still struggling to keep up with their rapidly expanding populations.

24
References

[1] Djahel, Soufiene, Nafaa Jabeur, Robert Barrett, and John Murphy. "Toward V2I communication

technology-based solution for reducing road traffic congestion in smart cities." In 2015 International

Symposium on Networks, computers and communications (ISNCC), pp. 1-6. IEEE, 2015.

[2] Jain, Vipin, Ashlesh Sharma, and Lakshminarayanan Subramanian. "Road traffic congestion in the

developing world." In Proceedings of the 2nd ACM Symposium on Computing for Development, pp.

1-10. 2012.

[3] Bacon, Jean, Andrei Iu Bejan, Alastair R. Beresford, David Evans, Richard J. Gibbens, and Ken

Moody. Using real-time road traffic data to evaluate congestion. Springer Berlin Heidelberg, 2011.

[4] Hurdle, V., E. Hauser, and Ph Fargier. "Effects of the choice of departure time on road traffic

congestion. Theoretical approach." Transportation and traffic theory 8 (1983): 223-263.

[5] Ata, Ayesha, Muhammad Adnan Khan, Sagheer Abbas, Gulzar Ahmad, and Areej Fatima. "Modelling

smart road traffic congestion control system using machine learning techniques." Neural Network

World 29, no. 2 (2019): 99-110.

[6] Arnott, Richard, and Kenneth Small. "The economics of traffic congestion." American scientist 82,

no. 5 (1994): 446-455.

[7] Yasin, Yasin J., Michal Grivna, and Fikri M. Abu-Zidan. "Global impact of COVID-19 pandemic on

road traffic collisions." World journal of emergency surgery 16, no. 1 (2021): 1-14.

[8] Jayapal, Cynthia, and S. Sujith Roy. "Road traffic congestion management using VANET." In 2016

International conference on advances in human machine interaction (HMI), pp. 1-7. IEEE, 2016.

[9] Verhoef, Erik T. "Time, speeds, flows and densities in static models of road traffic congestion and

congestion pricing." Regional Science and Urban Economics 29, no. 3 (1999): 341-369.

25
[10] Jayasooriya, S. A. C. S., and Y. M. M. S. Bandara. "Measuring the Economic costs of traffic

congestion." In 2017 Moratuwa Engineering Research Conference (MERCon), pp. 141-146. IEEE,

2017.

[11] Agbonika, Frederick. "Road Traffic Congestion and the quest for effective Transportation."

In NATIONAL ENGINEERING CONFERENCE AND ANNUAL GENERAL MEETING" CANAAN.

2011.

[12] Saroufim, Alain, and Elie Otayek. "Analysis and interpret road traffic congestion costs in Lebanon."

In MATEC Web of Conferences, vol. 295, p. 02007. EDP Sciences, 2019.

[13] Kiunsi, Robert B. "A review of traffic congestion in Dar es Salaam city from the physical planning

perspective." Journal of Sustainable Development 6, no. 2 (2013): 94.

[14] Redmon, Joseph, and Ali Farhadi. "Yolov3: An incremental improvement." arXiv preprint

arXiv:1804.02767 (2018).

[15] Mao, Qi-Chao, Hong-Mei Sun, Yan-Bo Liu, and Rui-Sheng Jia. "Mini-YOLOv3: real-time object

detector for embedded applications." Ieee Access 7 (2019): 133529-133538.

[16] Zhao, Liquan, and Shuaiyang Li. "Object detection algorithm based on improved

YOLOv3." Electronics 9, no. 3 (2020): 537.

[17] Li, Chong, Rong Wang, Jinze Li, and Linyu Fei. "Face detection based on YOLOv3." In Recent

Trends in Intelligent Computing, Communication and Devices: Proceedings of ICCD 2018, pp. 277-

284. Springer Singapore, 2020.

[18] Taheri Tajar, Alireza, Abbas Ramazani, and Muharram Mansoorizadeh. "A lightweight Tiny-

YOLOv3 vehicle detection approach." Journal of Real-Time Image Processing 18, no. 6 (2021):

2389-2401.

[19] Wang, Xiaolan, Shuo Wang, Jiaqi Cao, and Yansong Wang. "Data-driven based tiny-YOLOv3

method for front vehicle detection inducing SPP-net." IEEE Access 8 (2020): 110227-110236.

[20] Kim, Kwang-Ju, Pyong-Kun Kim, Yun-Su Chung, and Doo-Hyun Choi. "Performance enhancement

of YOLOv3 by adding prediction layers with spatial pyramid pooling for vehicle detection." In 2018

26
 15th IEEE international conference on advanced video and signal based surveillance (AVSS), pp. 1-

6. IEEE, 2018.

[21] Miao, Yan, Fu Liu, Tao Hou, Lu Liu, and Yun Liu. "A nighttime vehicle detection method based on

YOLO v3." In 2020 Chinese Automation Congress (CAC), pp. 6617-6621. IEEE, 2020.

[22] Zhao, Shuai, and Fucheng You. "Vehicle detection based on improved yolov3 algorithm." In 2020

International Conference on Intelligent Transportation, Big Data & Smart City (ICITBS), pp. 76-79.

IEEE, 2020.

[23] Zhang, Xun Xun, and Xu Zhu. "Moving vehicle detection in aerial infrared image sequences via fast

image registration and improved YOLOv3 network." International Journal of Remote Sensing 41,

no. 11 (2020): 4312-4335.

[24] Madakam, Somayya, Vihar Lake, Vihar Lake, and Vihar Lake. "Internet of Things (IoT): A literature

review." Journal of Computer and Communications 3, no. 05 (2015): 164.

[25] Gokhale, Pradyumna, Omkar Bhat, and Sagar Bhat. "Introduction to IOT." International Advanced

Research Journal in Science, Engineering and Technology 5, no. 1 (2018): 41-44.

[26] Van Kranenburg, Rob, and Alex Bassi. "IoT challenges." Communications in Mobile Computing 1,

no. 1 (2012): 9.

27