Adaptive Traffic Sensing Drone

Rajasekhar Chandana Venkateswara Rao Bathina Rajini Gudipudi

DepartmentofElectricalandElectronics DepartmentofElectricalandElectronics DepartmentofElectricalandElectronics
Engineering Engineering Engineering
V.R.SiddharthaEngineeringCollege V.R.SiddharthaEngineeringCollege V.R.SiddharthaEngineeringCollege
Vijayawada, India Vijayawada, India Vijayawada, India
[email protected] [email protected] [email protected]

Akshith Roy Kopuri Sri Jayanth Javvaji

DepartmentofElectricalandElectronics DepartmentofElectricalandElectronics
Engineering Engineering
V.R.SiddharthaEngineeringCollege V.R.SiddharthaEngineeringCollege
Vijayawada, India Vijayawada, India
[email protected] [email protected]

Abstract—A major problem in our rapidly changing society

is traffic congestion. The management of traffic using the congestion index ranking
conventional method is difficult due to the increase in the usage
of more private automobiles and the limited capacity of the road 80
network. The number of automobiles per household and on the
road increases along with the exponential growth in population. 60
As quickly as possible, a quick fix should be implemented to
minimize traffic congestion. The use of conventional techniques 40
may not be an effective and meaningful remedy for different
types of traffic congestion. The internet of things (IoT), machine 20
learning (ML), and artificial intelligence (AI) are currently
playing significant roles in the resolution of several real-world 0
issues. In order to find the best solutions for this problem, AI, Dec-19 Mar-20 Jun-20 Sep-20 Jan-21
ML, and IoT can be used. This essay aims to generate ideas for
drone-based traffic sensing. mumbai bengaluru
Keywords—Traffic, AI, IoT, ML ,Drone . new delhi pune

I. INTRODUCTION
Fig 1. Congestion index ranking in India in 2020
All over the world, India is the highest congestion country.
Traffic congestion in cities has become worse over the past Magnetic loop detectors were already employed to count
several decades, having a detrimental impact on everyday life the number of cars that pass over them. Systems for video
for people as well as limiting steady economic and societal monitoring that rely on vision have a variety of advantages
growth. Travel time, air pollution, and financial losses all rise over prior technologies. In these kinds of systems, a
as a result of traffic congestion. Governments work harder considerably wider range of traffic characteristics, including
than ever to control and relieve traffic congestion, but the vehicle classes, lane changes, parking spaces, etc., may be
effort is challenging due to the complexity of the issue. It is measured in addition to vehicle counts.
very challenging to predict. The dynamic and linked nature of
traffic congestion also reflects the complexity of the situation.
It can spread to nearby road segments from a busy road
CONGESTION
segment. Fully autonomous analysis of traffic congestion is
challenging to do due to its complexity.
PERCENTAGE AROUND
Fundamental data for a range of transportation initiatives,
THE WORLD
from transportation planning to contemporary intelligent
transportation, include traffic counts, speed, and vehicle 36%
classification. Even yet, sensors are still used by "Traffic 28%
36%
Monitoring" and "Information Systems" connected to vehicle PERCENTAGE 53%
56%
classification to estimate road conditions. We chose 2019 data 63%
due to the COVID 19 pandemic. There aren't enough cars on 65%
the road to rely on data. NEW YORK VANCOUVER PARIS
0% 20% 40% 60% 80%
JAKARTA MOSCOW BOGOTA
MUMBAI

Fig 2. Congestion index ranking around the world

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

 II. LITERATURE REVIEW various levels of filtering. Images are converted to binary
Utilizing an open tool for image processing, the number of format by replacing each pixel with one of two values: white
cars is computed by first displaying various image (logical 1) if the pixel's brightness is equal to or greater than
conversions on the screen and then enclosing the box around the level, or black (logical 0) otherwise. The specified level
the vehicle in the provided picture. By comparing the four must fall between [0, 1].
aspects of the image that is provided as input, they may use
the MatLab tool to determine the vehicle's density. They can
replicate the output of the four provided input photos;
however, this cannot be utilized in real-time applications since
the software is extremely inefficient and not freely available
[1].
Intelligent Traffic Signal System Based on Density The
optimization of a city's traffic light controller utilizing IR Fig 3. original frame taken from video
sensors and a PIC microprocessor. It was successfully
attempted to reduce the possibility of traffic congestion caused C. Image Enhancement
by traffic signals by adopting this system. The number of Digital pictures are modified throughout the image
passing vehicles during a specific time period on the road enhancement process to make them more suited for display or
determines the traffic density range, and a microcontroller in-depth examination. For instance, by removing noise, the
determines the traffic signal delays for the following recording item will be simpler to detect. With the exception of image
interval based on the vehicle count. Through connection rode method, image enhancement is the same as that utilized
between the microcontroller and the computer, the recorded in image detection and video detection. The function
data may be downloaded to the computer [2]. designated for this operation is used to first remove tiny linked
components and objects from the binary picture. These items
The projectis a simulation of a four-way intersection in a are below the stated threshold in terms of pixel count. For
real-world situation. The RFID idea is utilized to turn the instance, if we set the threshold to 10 pixels, any item with a
ambulance's lane green and so offer a stoppage-free path for size lower than 10 would be removed. One of the most crucial
the ambulance in the first section, which focused on problems and challenging prehending approaches is removing the
faced by ambulances. The second section, which focuses on image's noise, but after this is done, the task will get simpler.
issues with Priority vehicles, employs IR transmitters and
receivers to turn the vehicles' lane Green and therefore reduce The second phase enlarges and smoothed out the white
traffic congestion. IR transmitter and receiver are utilized to regions while filling in the black sections close to borders and
offer dynamic traffic management in the third segment, which perimeters. Dilate process has two parameters: one to define
is focused on Traffic density control. The project is a the kind of shape you want to draw and the other is a size of
simulation of a four-way intersection in a real-world situation matrix containing 1s and 0s; the location of the 1s defines the
[3]. neighborhood for the morphological operation.
Hence, we must identify the sensing drone from the The first parameter is the image from the stage before it,
module. AI, ML, and IoT may be utilised to discover the best and the other is defined by another function by creating a flat
solutions to this challenge. This project tries to spark concepts structuring element with the specified neighborhood. The
for traffic sensing using drones. central component of the matrix is its origin or center.
III. THE TRAFFIC SENSING MODULE D. Vehicle Enhancement
There are several approaches to reduce traffic congestion, The video analysis includes the identification of moving
but none of them has been very successful. Real-time traffic vehicles. It may be applied to a variety of situations, including
data is collected using a camera mounted on a drone that is video surveillance, traffic monitoring, and person tracking.
flown through the traffic in order to provide the percentage of Frame difference is one of numerous motion segmentation
traffic. The data is then sent back to a computing device where methods. The frame difference approach is simple to use and
it is analyzed, computed, and converted into percentage has a low computing complexity; because the difference
figures before being made available to the general public to between the current frame and the reference frame is greater
check on traffic details than the threshold, it is regarded as a moving vehicle. Another
approach Even when the camera moves, the optical flow
A. Capturing Image. approach can detect moving objects, but it takes more time
We can take the recorded picture from a live camera that due to its computational complexity and is particularly noise-
can record an image every ten seconds. However, for my sensitive.
study, we will take the video, split it into frames, and capture
a frame around every few seconds. If we utilize the camera,
we should mount it in a stable location to prevent vibration.
B. Foreground Detector
The Foreground Detector, which has a significant impact
on the filter and may identify the ground, is regarded as the
most crucial function in this code. Using Gaussian Mixture
Models, the foreground detector finds the foreground (GM-
M).Additionally, Foreground Detector converts the picture Fig 4. original frame taken from video
type from "RGB" to "Gray" and then "Binary" and performs
E. Vehicle tracking V. RESULTS
Vehicle tracking involves continuously identify the
On a laptop with an Intel Core i5 (2.4 GHz) CPU and 8GB
detected vehicle sequence and is done by specifically marking
the boundaries around the detected vehicle. Vehicle tracking RAM, we executed the algorithm. iNTEX IT-305 WC camera
is a challenging problem. Difficulties in tracking vehicles can is included.We did test flights in city traffic and flew the drone
arise due to abrupt vehicle motion, changing appearance for ten minutes in a steady position above a highway. Most
pattern of the vehicle, and vehicle to vehicle cars can be tracked and classified by the system with success.
However, due to the vehicle speeds, tests conducted in city
In my paper we use also from Foreground Detector Blob traffic were superior to those conducted on highway data. Due
Analysis function. This function detects the vehicles and then to their speed, two-wheelers made up the majority of the
from the bounding box we get the size of the detected vehicles. vehicles the system missed on highways. Figures display
After that, we draw a rectangle around the detected vehicle. some of our system's outcomes.

F. Data Transmission
1. Camera: Camera is used to take the video recording.
2. Raspberry pi 4b:It is used to transmit the data from the
camera to user interface
3. Data collecting and Traffic sensing unit: it holds the data
in the memory or USB services.
4. Data transmitting unit: using5GHZradio signal the data
is transmitting through to display.

Fig7. detecting vehicles by drone module

Data
ca Rasp collec Transm
m berr ting itting use
and data r
er y pi traffic
a 4B sensi
ng

Fig 5. Block diagram of data transmitting

IV. UAV MODULE

Processed data from drones. To produce a sorted, useful,
and coherent data collection, hundreds or even thousands of
photos are recognized, grouped, and/or stitched together, Fig8 .classification of detected vehicle
Measured. Drone data is frequently processed to create data
products, such a map or computer model, that may be
precisely geolocated and/or verified. VI. CONCLUSION
In high traffic areas where the reason of the traffic cannot
be detected on foot, this method of traffic detection, which can
be employed in places with frequent traffic and is quite
successful, is very useful. We have shown a vision-based
system for the accurate identification and counting of moving
automobiles on highways in order to efficiently gather the
data. The code has been compiled, and the adaptive traffic
sensing drone is producing results.

Fig6. Complete Quadcopter under its development phase.

 REFERENCES

[1] Neeraj K. Kanhere, Stanley T. Birchfield, Wayne A. Sarasua, Tom C. Intelligent Transportation Systems, Vol. 3, No. 1, March 2002, pp. 37–
Whitney, Real-time detection and tracking of vehicle base fronts for 47.
measuring traffic counts and speeds on highways", Transportation [11] Daily D. Cathy F. W. and Pumrin S. An Algorithm to Estimate Mean
Research Record, No. 1993. (2007). Traffic Speed Using Uncalibrated Cameras. IEEE Conference for
[2] Surendra Gupte, Osama Masoud, Robert F. K. Martin, and Nikolaos P. Intelligent Transportation Systems, 2000, pp. 98–107.
Papanikolopoulos, Detection and Classification of Vehicles, in Proc. [12] Koller D. Weber J. and Malik J. Robust Multiple Car Tracking with
IEEE Transactions On Intelligent Transportation Systems, Vol. 3, NO. Occlusion Reasoning. European Conference on Computer Vision,
1, March 2002. 1994, pp. 189–196
[3] A. J. Kun and Z. Vamossy, "Traffic monitoring with computer vision," [13] Koller D. Dandilis K and Nagel H. H. Model Based Object Tracking in
Proc. 7th Int. Symp. Applied Machine Intelligence and Informatics Monocular Image Sequences of Road Traffic Scenes. International
(SAMI 2009) . Journal of Computer Vision, Vol. 10, No. 3 1993, pp. 257–281.
[4] D. Beymer, P. McLauchlan, B. Coifman, and J. Malik, “A real-time [14] Haag M. and Nagel H. Combination of Edge Element and Optical Flow
computer vision system for measuring traffic parameters,” in Proc. Estimate for 3D-Model-Based Vehicle Tracking in Traffic Image
IEEE Conf. Computer Vision and Pattern Recognition, Puerto Rico, Sequences. International Journal of Computer Vision, Vol. 35, No. 3
June 1997, pp. 496–501. 1999, pp. 295–319.
[5] K. P. Karmann and A. von Brandt, “Moving object recognition using [15] Ferryman J. M. Worrall A. D. and Maybank S. J. Learning Enhanced
an adaptive background memory,” in Proc. Time-Varying Image 3D Models for Vehicle Tracking. British Machine Vision Conference,
Processing and Moving Object Recognition,vol. 2, V. Capellini, Ed., 1998, pp. 873–882.
1990. [16] Schlosser C. Reitberger J. and Hinz S. Automatic Car Detection in
[6] Thou-Ho Chen, Jun-Liang Chen, Chin-Hsing Chen and Chao-Ming High Resolution Urban Scenes Based on an Adaptive 3DModel.
Chang, "Vehicle Detection and Counting by Using Headlight EEE/ISPRS Joint Workshop on Remote Sensing and Data Fusion over
Information in the Dark Environment" in IEEE 2007 International Urban Areas, Berlin, 2003, pp. 98–107.
Conference on Intelligent Information Hiding and Multimedia Signal [17] Kamijo S. Ikeuchi K. and Sakauchi M. Vehicle Tracking in LowAngle
Processing IIHMSP07, Kaohsiung, Taiwan, Nov. 26-28, 2007. pp. and Front View Images Based on Spatio-Temporal Markov Random
519- 522. Page 36 of 37 Fields. Proc., 8th World Congress on Intelligent Transportation
[7] Nilesh J Uke and Ravindra C Thool, "Cultivating Research in Systems (ITS), 2001.
Computer Vision within Graduates and Post-Graduates using Open [18] Beymer D. McLauchlan P. Coifman B. and Malik J. A Real Time
Source", International Journal of Applied Information Systems 1(4):1- Computer Vision System for Measuring Traffic Parameters. IEEE
6, February 2012. Conference on Computer Vision and Pattern Recognition, 1997, pp.
[8] Klein L. A. Sensor Technologies and Data Requirements for ITS. 495– 501.
Artech House, Boston, Mass., 2001. [19] Saunier N. and Sayed T.T. A Feature-Based Tracking Algorithm for
[9] Middleton D. Gopalakrishna D. and Raman M. Advances in Traffic Vehicles in Intersections. Proc., 3rd Canadian Conference on
Data Collection and Management. White paper. Prepared by Battele, Computer and Robot Vision, June 2006.
Texas Transportation Institute, and Cambridge Systematics, Inc., for [20] Chachich A. Pau A. Barber A. Kennedy K. Oleiniczak E. Hackney J.
Office of Policy, FHWA, Washington, D.C., Jan. 2003. Sun Q. and Mireles E. Traffic Sensor Using a Color Vision Method.
[10] Gupte S. Masoud O. Martin R. F. K. and Papanikolopoulos N. P. Proceedings of the International Society for Optical Engineering, Vol.
Detection and Classification of Vehicles. IEEE Transactions on 2902, Jan. 1997, pp. 156–164