Traffic management in smart cities

 Rahul Gonela-1NH20CS079
 Kantham Thirumal Reddy – 1NH20CS103
 Jarugula Ganesh Chowdary-1NH20CS094

GUIDE :
Dr.B.Meenakshi Sundaram
Professor
Abstract

 Congestion in traffic is a serious issue . In existing system signal timings are fixed and
they are independent of traffic density,
 Large red light delays leads to traffic congestion.
 In this project, Video based traffic estimation system is implemented in which signal
timings are updated based on the vehicle counting.
 This system consists of Deep Learning module it detect the vehicle count of the current
system and sends to the traffic signal.
 Based on traffic density of Vehicle System will Predict the Traffic Congestion.
INTRODUCTION

 Almost all urban cities in the world use traffic light signals to control the traffic on the
roads.
 Different types of traffic light control systems are developed which are vehicle actuated
lights and static traffic lights.
 But their traffic lights timing are fixed and switching patterns are also predefined in the
system it is independent of traffic conditions for the different lanes and they are not
changing with real time data.
OBJECTIVES

 Objectives Eliminate of frequent traffic disconnect of Road network.

 Act as Road side Unit as server.
 Implementation of Ambulance Priority system.
EXISISTING SYSTEM

 In existing system human and Automatic based traffic light Control system system only
mostly in use.
 Also some of the sensor based Traffic monitoring also used.
 These techniques are having huge drawbacks.
PROPOSED SYSTEM

 This system proposes a new system for predicting the traffic Density by image processing
using AI module.
 A camera will be installed alongside the traffic light.
 It will capture image sequences.
 The image sequence will then be analyzed using digital image processing for vehicle
detection, and according to traffic conditions on the road, traffic Density can be
Estimated.
PROPOSED SYSTEM

 This system employs YOLOV5 and AlexnetNet v_{3} Convolutional neural network pre-
trained model to accurately detect the number of vehicles present on the road, Average
Vehicle Area and identify emergency vehicles in real-time.
 Using this information, this system can dynamically adjust traffic signals and reroute
vehicles to minimize congestion and ensure priority access for emergency vehicles.
DISADVANTAGES

 That human based system need huge manpower in the form of traffic Police.
 Less accuracy of processing sensor signals.
 Prone to Sensor and Man Made errors which leads to system efficiency drop.
 SURVEY AUTHOR YEA EXPLANATION
R

LITERATURE SURVEY 1: J,Redmon and 2020 • Deep convolutional Network

Yolo9000: A.Farhadi • Colour point
SURVEY Better,faster,stronger • Pixel count

SURVEY 2: Nilakorn 2019 • Reference Line Model with

A computer vision based Seenouvong Gaussian mixer
vehicle detetection and Ukrit • Grey scale countuing.
counting system Watchareer
ueta
SURVEY 3: J.Redmon 2018 Majority pixel Count
“you only look S.Divvala Colour and Hidden Markov Model.
Once:Unified,real-
Time object detection.
IMAGE ACQUISITION MODULE

 Image Acquisition Module Image acquisition is done by using an external Video.

 In this project, the operating system used is LINUX which is an open source software that
is prone to frequent changes.
IMAGE PREPROCESSING MODULE

 Image preprocessing functions are imported from Python OPEN CV libraries and is
included in final python program.
 This will automatically process the image when the program is invoked.
VEHICLE DETECTION
OBJECT DETECTION MODULE

 Among the key features of an image i.e. edges, lines, and points, we have used edge in
our present work.
 Here we are using object detection method for image matching.
 Object detection methods locate the pixels in the image that correspond to the shape of
the objects seen in the image.
 The result is a binary image with the detected Object.
VEHICLE COUNTING MODULE

 Vehicle Count Initially reference image of an empty road is stored in memory.

 The captured images at the four lanes are then compared with the reference image to
determine the vehicle density using Object Detection Module
 HARDWARE
REQUIREMENTS

 System : Pentium Dual Core

 Hard Disk : 120 GB.
 Monitor : 15” LED
 Input devices : Keyboard,Mouse
 Ram : 4 GB
SOFTWARE REQUIREMENTS :
 Python Language
 Anaconda IDE
ADVANTAGES

 The process is in real time so in can able to detect from the video stream from live
Camera.
 The implementation of this approach runs at 30-40 frames per second, so that it can detect
the vehicle very quickly.
 It uses low power processor with 2.4 GHz by using that we can able to achieve low power
operation.
Light Control Module

 After vehicle detection procedure both reference and real time images are matched and
traffic lights can be controlled based on percentage of matching.
 If the matching is between o to 10% - green light is ON for 90 seconds.
 If the matching is between 10 to 50% - green light is ON for 60 seconds. If the matching
is between 50 to 70% - green light is ON for 30seconds.
 If RF Signal Received from Ambulance it will Open the Green Signal Always for
Ambulance Passing
FUTURE ENHANCEMENT

 Higher versions of raspberry pi can be used for reducing the processing time
 Timer system can also be incorporated
 While determining vehicle density if the number of vehicles on any particular side is
always less, then each time the lane will be neglected and the waiting time of vehicle on
that particular lane will be high.
 For future expansion this has to be considered
CONCLUSION

 This project tries to introduce a new method for traffic Estimation based on vehicle
density, which is found to be very efficient method.
 As opposed to digital image processing using bulky computers, this method use AI
module, a microcomputer for image processing which greatly reduces the space and
processing time.
 To determine traffic density different object detection techniques can be used.