FINAL PROJECT PRESENTATION

A I Based Traffic Control And Anomaly Detector

Guide : Mr.Sudeer V R GROUP 8

Assistant Professor
Department Of ECE RAHUL R
Co Guide : Mrs.Seena R
Assistant Professor JITHU S A
Department Of ECE VAISAKH P SUGATHAN
 ABSTRACT

Traffic control has been a crucial problem since decades.

An adaptive live traffic control system that works effectively in conjunction

with various scientific methods.

Artificial Intelligence, Machine Learning and On-Chip computerisation

The traffic module changes the duration of signal lights by analysing the
traffic density

Surveillance ,emergency exits and performance enhancement is made

possible.

1
 INTRODUCTION

• Traffic Lights have become an integer part of human’s

day-to-day life.

• One way to improve traffic flow and safety of the

current transportation system is to apply automation
and intelligent control methods to roadside
infrastructure .

• The project aims in designing and implementing, a

prototype of traffic light controller which is controlled
smartly.

2
 Cont ….
INTRODUCTION

• Adaptive live traffic management system

• In reference to Intelligent Transportation System [ITS]
• Replacing conventional traffic system
• Development in 4 major areas
Density Measurement
Emergency Vehicle Detection
Smart Weather Detection
Violation Detection System

3
 LITERATURE SURVEY

• Idea of traffic regulation came in 1956

• Proposed by ATMS [ Advanced Traffic Management System ]
• Started with development of roads in structural wise

4
 Cont ….
LITERATURE SURVEY

• Intelligent Road Traffic Control System for Traffic Congestion :

A Perspective
[ Pallavi A Mandhare, Vilas Kharat & C.Y.Patil ]
• Present usable models and algorithms
• Basic Traffic Controls to Intelligent Signal Timing
Systems using Neural Networks
• Algorithm varies from infrastructure to vehicle
communication to dynamic programming with vehicle
parameters to IoT implementation
• Briefs out the history of ITS

5
 LITERATURE SURVEY

• Intelligent System in Traffic Response with Digital Image Processing in ITS

(Ali Tourani [2010] )

• Application of DIP in Vision BasedITS

• Utilise Roadways or in-vehicle cameras
• Applications : Automatic License Plate Recognition, Traffic Flow
Estimation,Speed Measurement, Vehicle Type Identification were the
Applications

6
 LITERATURE SURVEY

• A Dynamic Road Incident Information Delivery Strategy to Reduce Urban

Traffic Congestion
[ Vladimir Uremovic (2010) ]
• Incident detection and blocks road to reduce incident induced
road block
• Time-Dependent shortest path algorithms used later
• Traffic Density Measurement was also mentioned as an
incident parameter
• Traffic flow transferring real time information

7
 GOALS

• TRAFFIC SIGNAL MODULATION

• EMERGENCY VEHICLE DETECTION

• VIOLATION DETECTION

• WEATHER DETECTION

8
 SOFTWARES USED

PYTHON

• Python is an object-oriented, high-level programming language,it supports

the use of modules and packages in programming.

• Python has a huge set libraries which can be easily used for machine
learning.

• As its platform undefended its easy for us to run this program on any
operating system

• This project is completely coded with python programming language

{expect Arduino programs}

9
 SOFTWARES USED

Open CV

• Open Source Computer Vision Library) is a library of programming functions

used for real-time computer vision, developed by Intel.

• The OpenCV library contains hundreds of functions that support the

capture, analysis, and manipulation of visual information fed to a computer
by webcams, video files, or other types of devices.

• OpenCV captures the images from a camera pointed at each of the roads
and sends it for further image processing

• We used open cv especially in density measurement and violation detection

algorithms

10
 SOFTWARES USED

YOLO
• YOLO is a convolutional neural network based model that detects objects in
real time. It is based in dark-flow and can detect over 9000 different objects
with 70% accuracy.

• YoLo algorithm is applied to the images captured from the road and with the
help of tensor flow it Calculates the number of vehicles in the road.

• Biggest advantages:

• Speed (45 frames per second).


Network understands generalized object representation.

• Faster version (with smaller architecture) — 155 frames per sec but
is less accurate. . Open source 11
 SOFTWARES USED

TENSER FLOW

• TensorFlow is an open source software library for numerical computation

using data-flow graphs. It was originally developed by the Google. .

• TensorFlow can train and run deep neural networks for handwritten digit
classification, image recognition, word embeddings, recurrent neural
networks etc.

• Convolutional networks are applied to the image using tensor flow to identify
the type and number of vehicles and then allocates time slots for each road.

• In this projects we used tensor flow as the backend platform to run all the
programs.

13
 Cont ….
DENSITY MEASUREMENT ALGORITHM

• In the beginning we decided to use on chip computer and web

cam .
• But after the implementation. We faced may challenges such as
• Slow data processing by one chip computers
• Low resolution for images
• Web cam require high bandwidth for serial port
communication
• Threading is not possible .

14
 Cont ….
DENSITY MEASUREMENT ALGORITHM

• Designed calculate no of vehicle on corresponding road

• Using 4 wireless camera taking inputs from different road
• Calculating no of vehicle from that image
• As per the density we are assigning duration of time
• Rather than taking video inputs for calculation we use
snapshots in special manner to produce fast and efficient
manner

15
 Cont ….
DENSITY MEASUREMENT ALGORITHM

Density Algorithm

Port connection
Step 1 establishing with Camera
and Arduino

Turning on all red initialisation of all  traffic

Step 2 light lights 

Step 3 Taking Snapshot of

Track 1

If there is
Step 4 vehicle 

After calculating Number of

Step 5 vehicle Assigning the timer for
track 1 vehicles

Taking Snapshot of
Step 6 Track 2

If there is
Step 7 vehicle 

After calculating Number of

vehicle Assigning the timer for
Step 8 track 2 vehicles

After step 8 , repeat from step  3 for track 3.

Logical Flow Chart of density algorithm

16
 Cont ….
DENSITY MEASUREMENT ALGORITHM

INPUT OUTPUT

17
 EMERGENCY VEHICLE DETECTION

• Priority for emergency vehicle is very import in traffic

management system
• Identifying a emergency vehicle after entering the traffic area
using cameras is not efficient way of management
• So we studied about the characters of emergency vehicle we
able to find out all emergency vehicle use alarms with sound
between 120-140db.
• Using that feature we designed emergency vehicle detection
algorithm

18
 Cont ….
EMERGENCY VEHICLE DETECTION

• We allocated audio device(mic) in each track before 400-500

meters of traffic jam area .
• When an emergency vehicle pass through the mic will be get
detected and make the corresponding path cleared
• In order to avoid false sound recognition we provide certain
criteria in algorithm for ensuring the perfect detection
• After the given interval of time traffic will resume from where its
is interrupted .

19
 Cont ….
EMERGENCY VEHICLE DETECTION

Emergency Detection

Initialisation of audio 
equipments and  setting the
 sound meter limit to 220-230 dB

If the audio signals  come in given Waiting for Audio signal in that
range for 3-4 seconds continuously range

Detects Emergency
Vehicle

Turning ON green
light in corresponding
route

After the given time

period traffic light
back to normal
working algorithm

20
Logical Flow chart of emergency vehicle detection algorithm
 Cont ….
EMERGENCY VEHICLE DETECTION

INPUT OUTPUT

• Playing sound of emergency • Soon after the sound

detected
vehicle near the input mic
• Traffic lights of
corresponding road become
green and others become
red
• For given period of
time ,then come back to
normal working

21
 WEATHER DETECTION ALGORITHM

• As the climate change can directly affect the performance of

cameras in taking picture ,its will be difficult for calculating the
no of vehicle especially in rainy seasons
• Hence we use a weather detection sensor to sense the climate
change but measure the atmospheric humidity
• When there rainy or mist the sensor detects the climate change
• As per the detection regular density calculating algorithm will
shifted to special algorithm

22
 Cont ….
WEATHER DETECTION ALGORITHM

• Here rather than taking one snapshot we are taking

simultaneously no of snapshot
• Then calculating no of vehicle in each image ,after that taking
the average of value as the actual output
• With respect to that value program will work further.

23
 Cont ….
WEATHER DETECTION ALGORITHM

Weather Detection

Connecting Weather Detection sensor

with program through Arduino 

if change in weather Running the  regular

detects  Density measurement
Algorithm

Rain detection sensor with LM393 Comparator

Running Special
algorithm for density
measurement
24
 VIOLATION DETECTION ALGORITHM

• We designed this algorithm for detecting the violating vehicle

when the traffic lights are red
• But as the input is video it’s takes a lot of time to process and
classify objects in video .
• Thus we require a high end graphical processor to get real time
output.
• Rather than looking into live output we can collects images of
traffic rule violated vehicle from separate folder any time.

25
 Cont ….
VIOLATION DETECTION ALGORITHM

• Initially we coded the program to detect all vehicle in image with

opencv and YOLO weights.
• Later we bound a line to mention the crossing limit of vehicle in
road inside a image of road.
• So when the vehicle cross the line { if the traffic light is red }
program will capture the images of such vehicle and stores in
separate location.

Output 26
Input
 ADVANTAGES AND DISADVANATGES

ADVANTAGE DISADVANTAGE

• Up gradations results in immense increase in cost.

• ︎ Cost Efficient 


• ︎ Easy Implementation 
 • Possibility of weather detection is a question in severe conditions in this

hardware setup.
• ︎ Effective violation detection 

• Cannot rule out in junctions with more than 4 exits.
• ︎ Traffic Management using less human effort 


• Multi-lanes require cameras with high resolution and frame rate to capture large
• ︎ Greater scope for future up gradations 

area.

• ︎ Very little possibility of error 


•
• Accidents occurring within the field of view of camera and their response with it
is not 

deigned

• Violation detection implementation in this set up due to high graphic content

required 

separate SoC. 

27


 CONCLUSION

• Innovation and productivity coverges with social benefit.

• Wipes out major probelms of traffic infrastructure.

• Low cost and easily

• upgradable.

• Upgrading possibility opens with new technology in same field.

28
THANK YOU

28