“Automatic Traffic Management System Using YOLO”

[1]
Harshitha R, [2]Harshitha N R,[3]Kavitharani B S, [4]Madhura R S
[1]
harshiammu890@[Link],[2]harshitharanju14@[Link], [3]kavithagowdabs20@[Link],
[4]
madhurars2402@[Link],
[5]
Jamuna M
[5]
[Link]@[Link]
Associate Professor
Vidyavardhaka College of Engineering ,Mysuru,India

Abstract- Traffic congestion is one of the major counting and automatic signals Scheduling.
problem in today’s world, which is need to be Ambulances and fire brigades are usually
solved to improve traffic control and allowed by all the co-travelling vehicles on the
management. Vehicle flow detection appears to road on humanity grounds. But when it comes to
be an important part in today’s traffic huge cities with lot of traffic jam, these
management system. In this project, we propose emergency vehicles are not able to provide its
an automatic traffic management system for services on time. In this project, these emergency
vehicle detection and counting and automatic vehicles are detected and are allowed to move
signals Scheduling. The camera supplies video even in any traffic condition by keeping the rest
input to the processing engine. Initially the video of the vehicles on other ends in standby. The
will be streaming on all four roads of the traffic emergency vehicles will be given prominence in
circle. The values will be read frame by frame in all the traffic signals.
the streaming video of these roads. Camera sends
all the captured input images to the processing 2. LITERATURE SURVEY
engine and this works based on the neural
network. The traffic flow shows the traffic state TITLE:” Yolo Flow Real-time Object
in fixed time interval and helps to manage and Tracking in Video”
control the traffic especially when there is a
heavy traffic and will consider emergency YOLO is a fast, accurate object detector, in
vehicles like ambulance and fire brigades, giving TensorFlow. It contains publicly-available
them priority to go. dataset containing several million natural images.
Keywords-object detection, YOLO, signal WeusethePASCALVOC2007, a set of RGB
scheduling, deep learning, computer vision images labeled with bounding box coordinates
and class categories. The CNN learns high-
quality, hierarchical features automatically,
1. INTRODUCTION eliminating the need for hand-selected features. It
Solve the object detection and object
Traffic is a major issue in every city. The traffic classification tasks. This function simultaneously
mostly affects the emergency service vehicles penalizes incorrect object detections as well as
such as ambulances and fire brigades badly. considers what the best possible classification
Traffic congestion is one of the major problem in would be[1].
today’s world, which is need to be solved to
improve traffic control and management. Vehicle
flow detection appears to be an important part in TITLE :“ Smart Traffic Control System Using
today’s traffic management system. In this Image Processing”.
project, we propose an automatic traffic
This paper used for estimating the traffic and
management system for vehicle detection and
this method has been implemented using Image

Electronic copy available at: [Link]

Processing and Mat lab software. This is done by proportion of bounding coordinates error is
using the camera images captured from the increased. A new object detection method,
highway and videos taken are converted to the OYOLO (Optimized YOLO), is produced, which
image sequences. Each image is processed is 1.18 times faster than YOLO, while
separately and the number of cars has been outperforming other region-based approaches
counted. If the number of cars exceeds a specific like R-CNN in
threshold, warning of heavy traffic will be shown accuracy. To improve accuracy further, we add
automatically [2]. the combination of OYOLO and R-FCN to our
system. For challenging images in nights, pre-
processing is presented using the histogram
TITLE:“ “RFID and GPS based Automatic equalization approach. We have got more than
Lane Clearance System for Ambulance” 6% improvement in map on our testing set[4].
The lack of efficient traffic control and
management has many a times lead to loss of TITLE: “Evaluating State-of-the-art Object
lives due to ambulances getting stuck in traffic Detector on Challenging Traffic Light Data”
jams. To overcome this problem, we propose a
RFID and GPS based Automatic Lane Clearance Traffic light detection (TLD) is a vital part of
System for Ambulance. The focus of this paper is both intelligent vehicles and driving assistance
to reduce the delay in arrival of the ambulance to systems (DAS). General for most TLDs is that
the hospital by automatically clearing the lane in they are evaluated on small and private datasets
which ambulance is travelling, before it reaches making it hard to determine the exact
the traffic signal. This can be achieved by turning performance of a given method. In this paper we
the traffic signal, in the path of the ambulance, to apply the state-of-the-art, real-time object
green when the ambulance is at a certain distance detection system You Only Look Once, (YOLO)
from the traffic junction. The use of RFID on the public LISA Traffic Light dataset
distinguishes between the emergency and non- available through the VIVA-challenge, which
emergency cases, thus preventing unnecessary contain a high number of annotated traffic lights,
traffic congestion. The communication between captured in varying light and weather conditions.
the ambulance and the traffic signal post is done The YOLO object detector achieves an AUC of
through transceivers and GPS [3]. impressively 90.49 % for daysequence1, which is
an improvement of 50.32 % compared to the
TITLE: “An Object Detection System Based latest ACF entry in the VIVA challenge. Using
on YOLO in Traffic Scene” the exact same training configuration as the ACF
detector, the YOLO detector reaches an AUC of
We build an object detection system for images 58.3 %, which is in an increase of 18.13 % [5].
in traffic scene. It is fast, accurate and robust.
Traditional object detectors first generate
proposals. After that the features are extracted.
Then a classifier on these proposals is executed. 3. PROBLEM STATEMENT
But the speed is slow and the accuracy is not
The purpose of this project is to provide high
satisfying. YOLO an excellent object detection
priority for the emergency vehicles like
approach based on deep learning presents a
ambulances, fire brigades and VIP vehicles to go
single convolutional neural network for location
in traffic signals. In a busy schedule thousands of
and classification. All the fully-connected layers
vehicles passing through the traffic, it may not
of YOLO’s network are replaced with an average
always be possible for traffic police officer to
pool layer for the purpose of reproducing a new
schedule the signal instantly and usually the
network. The loss function is optimized after the

Electronic copy available at: [Link]

traffic signals will not do this automatically 5. SYSTEM ARCHITECTURE
based on the traffic strength.

4. METHODOLOGY

Initially the video will be streaming on all four

roads of the traffic circle. The values will be read
frame by frame in the streaming video of these
roads. Camera sends all the captured input
images to the processing engine and this works
based on the neural network. It is one of the
algorithm in machine learning and it has three
layers in it. First one is input layer, it stores
captured data. Second one is hidden layer, it
divides the images into regions and predict
boundary boxes and probabilities for each region,
this boundary boxes are weighted by the
predicted probabilities. Third one is output layer,
It has trained set of data. The counter value is set
to 60 seconds. When the emergency service
vehicle enters the non-green lane, that will be In our system architecture, first we will set timer
detected using YOLO and the signal wait timing for the camera. Then camera captures all the
is reduced and the vehicles in that lane are images coming through the lane. After that, it
allowed to move. After this the signal timing sends all captured input to the processing engine
works on the round robin [Link] the and it works based on the neural network. It is
emergency vehicle moves, the signal will become one of the learning algorithm in machine learning
green at the next lane and the vehicles are and it has three layers in it. First one is input
allowed to pass. This works in clock –wise layer, it stores an captured data. Second one is
[Link] the emergency vehicle was hidden layer, it divides the images into regions
allowed to pass, that lane’s signal waiting timing and predict boundary boxes and probabilities for
will be reduced to half of its actual timing i.e., each region, this boundary boxes are weighted by
[Link] the two or more emergency vehicles the predicted probabilities .Third one is output
are encountered at different lanes, then it works layer ,it has trained set of data. If the captured
on “First come first serve” method. data matches with ambulances or fire brigades
,automatically traffic light releases an green
signal color to the corresponding lane otherwise
4.1 DESIGN depending on the no. of vehicles it changes the
traffic signal color to green .

6. CONCLUSION

This system is a video based traffic controlling

system which is a real time application.
TensorFlow implementation of ‘YOLO: Real-
time object detection’ (train and test). This
project helps to manage and control the traffic
especially when there is a heavy traffic.

Electronic copy available at: [Link]

Emergency vehicles like ambulance and fire
brigades are given first priority to move. [4] Jing Tao ,Hongbo Wang ,Xinyu Zhang ,Xiaoyu
Li , Huawei Yang. “An Object Detection
7 . REFERENCES System Based on YOLO in Traffic Scene”

[5] Morten B. Jensen, Kamal Nasrollahi, and

[1] Konstantine Buhler, John Lambert, Matthew Thomas B. Moeslund Section of Media
Vilim “YoloFlow Real-time object Tracking Technology. “EvaluatingState-of-the-
in Video” CS 229 Course Project. artObjectDetectoronChallengingTrafficLight
[2] Prashant Jadhav, Pratiksha Kelkar, Kunal Data”,Nov 17,2017.
Patil, Snehal Thorat. ”Smart Traffic Control
System Using Image Processing”. Bachelor
of IT, Department of IT, Theem College Of
Engineering, Maharashtra, India.
[3] Rashmi Hegde, Rohith R. Sali & M. S.
Indira.” RFID and GPS based Automatic
Lane Clearance System for Ambulance”.

Electronic copy available at: [Link]