IOT Early Pedestrian Detection for Driving - Project Requirements Analysis and Specification

Running head: RESEARCH METHODS AND PROJECT DESIGN
MN691 Research Methods and Project Design
Project Title: IOT Early Pedestrian Detection for Driving
Assignment-1 Group report
Project Requirements Analysis and Specification
By
Rajasekhar Polaka- MIT173408
Manisha Kolle- MIT180076
Praneeth Kumar Desai Karanam- MIT175014
Chandrahas Reddy Bumireddy- MIT174741
Madhu Kumar Vattepu- MIT173378

1
RESEARCH METHODS AND PROJECT DESIGN
Table of Contents
1. Introduction................................................................................................................2
2. Problem Domain & Research Question.....................................................................2
Research Question......................................................................................................3
3. Background & Project Objective...............................................................................3
3.1. Aim & Objectives...............................................................................................3
3.2. Literature Review................................................................................................3
4. Project Requirements Analysis & Specification......................................................11
Table of weekly activities............................................................................................12
Responsibilities and roles of team members for assignment 2................................12
Gantt chart....................................................................................................................14
Project design...............................................................................................................18
Project individual research approach...........................................................................20
Budget..........................................................................................................................24
Responsibilities and roles of team members for next part of project.......................25
Research methods that are to be used in the next stage of the project.....................29
Flowchart of working of the system.........................................................................32
Working of the Microsoft Azure for storing data....................................................32
Prototype of the project................................................................................................33
Conclusion....................................................................................................................34
Limitations...................................................................................................................35
Future work..................................................................................................................35
Bibliography.................................................................................................................36
Appendix......................................................................................................................39
Appendix IV.............................................................................................................39
Appendix V..............................................................................................................39
Appendix VI.............................................................................................................40

2
RESEARCH METHODS AND PROJECT DESIGN
1. Introduction
The human errors, speeding and environmental changes are causes for road accidents.
In the current technology-driven world, most of the companies are depending on technology
to reduce human errors and efforts and increase production capacity and quality. With
emerging technologies, road accidents can be controlled and pedestrians and drivers safety
can be improved. Emerging technologies like IoT, Mobile phone with Data, GPS and Display
monitor for surrounding car situation are capable to monitor the driving environment. Internet
of Things (IoT) refers to the scenario where computing and network connectivity are moved
to sensors and objects. These devices and sensors become capable to exchange data without
human involvement. IoT technology can avoid road accidents; the sensors fixed to vehicles
sense the surroundings provide intimations to the driver through alert systems and voice
messages. IoT avoids road accidents and damage to vehicles; this ensures comfortable and
safe driving. GPS traffic flow notifications, shape-based detection systems, video-based
detection, and multi-cue vision systems are a few other technologies used to prevent road
accidents. IoT device i.e. Early Pedestrian Detection for Driving” device is easy to remove
and place in cars. Two major cases where IoT devices are used to avoid the accidents are
turning or reversing cars. In addition, IoT and other technologies are capable to detect the
breaker and conditions in the road; these improve road safety by reducing the road accidents.
Major sections discussed in the report are; reasons for increased road accidents, various
technologies used to detect the pedestrian, literature review and project requirements or
specifications to implement IoT to avoid road accidents. In requirements specifications
sections the hardware and software components required for the development and
implementation of Early Pedestrian Detection for the driving device. The literature review
discussed the various methods to detect the pedestrian crossings, Detection of Pedestrian
Crossing for Safe Driving, Pedestrian detection using the moving camera, Detecting
pedestrian using convolutional neural networks and A Multi-Sensor Fusion System for
Detection and Tracking pedestrians.
2. Problem Domain & Research Question
In this technology world, day by day new cars are invented and these are being
employed by people. As more people are using cars the traffic and accidents are also
increasing. Major causes of road accidents are over speeding, distractions to drivers, drunken
driving, red light jumping and avoiding safety gears. It is identified using emerging

3
RESEARCH METHODS AND PROJECT DESIGN
technologies the road accidents can be reduced and driver and pedestrian safety is improved.
The main purpose of this document is to identify and discusses how emerging technologies
are used to avoid accidents and improve safety. This report provides a clear explanation about
technologies architecture and functioning.
Research Question
What are possible benefits the Internet of Things (IoT) offer to vehicle drives
and how it reduce the road accidents. What are the biggest issues associated with the IoT.
3. Background & Project Objective
3.1. Aim & Objectives
Aim
The main aim of this report is to explain how Early Pedestrian Detection fordriving
device avoid road accidents and improve the safety of drivers.The functioning and possible
benefits of the device will be clearly explained.
Objectives of the current report are
 To conduct a literature review and identify the emerging technologies that are capable
to avoid road accidents and improve the safety.
 To explain the functioning of IoT and other emerging technologies. So that the drivers
can easily understand the concept adopt them.
 To identify and explain the hardware and software components required for the
implementation of Early Pedestrian Detection for the driving device.
3.2. Literature Review
Utilizing various methods to detect pedestrian crossings
In the aspect of the autonomous vehicle, finding pedestrian motions is the essential
thing to help in reducing road accidents. Because today many accidents occur due to the
sudden pedestrian’s crossings on the roads. For this reason, pedestrian safety has to become
an important aspect and safety components are emerged at the market to reduce road
accidents and to ensure safe driving. Some of the safety components such as GPS traffic flow
notifications, shape-based detection systems, video-based detection, multi-cue vision
systems, tracking systems, and alert systems are helping car users to prevent the road
accidents [1]. In addition, Advanced Driver Assistance Systems and intelligent vehicles are

4
RESEARCH METHODS AND PROJECT DESIGN
helping the vehicle users to recognize the surroundings. But in the context of machine vision,
recognizing the pedestrian’s sudden movements or crossings are still a difficult job. For this
reason, it is essential to build a system that must have the features to capture image or videos
of human motions on roads. To detect or recognize the motions of pedestrian, visible,
ultrasonic, and sensors should be used. Various programmed frameworks are provided to the
drivers to enhance the driving conditions and to reduce road accidents. To capture the picture
of pedestrians feature methods were used such as Covariance matrix and HOG Descriptors.
HOG is enhanced the performances of detection of pedestrians and HOG features are used in
the automated pedestrian detection systems to capture motion patterns of pedestrians. There
are five steps were majorly performed in the automated pedestrian detection systems such
steps are; obtain image and videos, computing motion vectors and extracting the information
with the help of HOG features, feeding this feature to discriminate the human and testing the
videos and motion patterns of pedestrians.[1]. The Heo et.al stated that there is a number of
road accidents occurred between pedestrian and vehicles due the late night and due to the
failure of detection of pedestrian crossings. The Adaptive Boolean Map Based Saliency
(ABMS) is used to recognize the pedestrian movements and this method was used to detect
conspicuous regions and this method was derived from the Boolean map theory. The You
Only Look Once (YOLO) is another classifier method that is used to classify the pedestrian
locations on the roads [3]. On the other hand, driver assistance system (DAS) is gaining the
popularity which is mainly used to obtain the wide view image and DAS has various
components such as sensors, radar, and image sensor to provide visual information to the
vehicle user so that pedestrian locations are easily detected. The image sensor of the DAS has
a special feature to capture the visual data of the objects that are presented at the vehicle
surroundings [4]. Wireless vehicle alert and collision prevention systems are also widely used
for reducing the vehicle accidents and these systems are integrated with the SMS alert system
and it detects the crashes between the pedestrian and vehicles and sends the SMS to the
relevant GSM details [5]. There are other methods have been introduced for detecting the
pedestrian movements on the roads [6]. The Convolutional Neural Networks (CNN) to detect
the pedestrian locations and it uses the Histogram of Oriented Gradients (HOG) features to
easily recognize the motion patterns of the pedestrians on the roads.
Detection of Pedestrian Crossing for Safe Driving
Early pedestrian detection for driving device helps drivers to escape from the crash
with the pedestrians while crossing around the cars during turning or reversing the car. SPCs

5
RESEARCH METHODS AND PROJECT DESIGN
strategy is mainly used for identifying virtual reference lines which are connected to street
detachment without selecting screening data. SPC system calculation is carried out by
exploiting Kalman filtering for tracking the pedestrians. The system effectively applied in
KMU SPC dataset where it determines the data with the use of FIR camera [1]. As a result of
these factors, SPC discovery is mainly used for calculating the recursive functions which a
linear quadratic estimation is called as Kalman filtering. The LBP extraction occupies centre
pixel and the threshold value of around 8 pixels. The process of thresh holding the binary
strings and storing the output decimal value of LBP array is repeated for each pixel of input
image [2]. The performance is based on the results of true positives and negatives, false
positives and negative. The system shows a higher TP rate for two caution and normal types
of SPC. So, therefore firstly detection of the pedestrian is done and SPC prediction is carried
out by Kalman filtering for object tracking and it generates approximately 13.58 seconds of
processing time and it generates false positive results in SPC detection during a rapid change
in vehicle direction [3]. Pedestrian detection has become one of the hottest topics in
intelligent traffic system because of its potential applications in driver assistance and
automatic driving.
Fast Pedestrian Detection and Dynamic Tracking for Intelligent vehicles
CamShift algorithm combined with extended Kalman filtering and it is mainly used
for pedestrian dynamic and as a result of these factors, smartphone-based V2V cooperative
warning system is developed to share useful detection results within blind spots [4]. For
future automated driving functions, it is necessary that intentions and future movements of
vulnerable road users in urban traffic scenarios. The given file mainly focuses on the
behavior of pedestrians at crosswalks. Relevance determination algorithm is mainly used to
identify the meaningful features within feature-space. It is mainly used for future automated
vehicles and advanced driver assistance systems. Reliable detection of pedestrians from a
moving vehicle is the key to preventing or mitigating accidents involving these most
vulnerable traffic participants [5]. Pedestrian detection is of particular interest to the
automotive domain, where an accurate estimation of a pedestrian’s position is the first step
towards reliable collision avoidance systems. Detected pedestrians in the vicinity of a vehicle
can be displayed to the driver in an intuitive fashion by utilizing a Bird’s Eye View approach.
To obtain the images from four calibrated cameras surrounding the vehicle are combined and
back-projected to the ground plane [6]. The EU-funded project Autonomous Valet Parking
and Charging aims to develop an autonomous electric vehicle that can move in parking lots

6
RESEARCH METHODS AND PROJECT DESIGN
and reach re-charging spots. Cameras are the main sensor of the V-Charge car platform
algorithms employ four fisheye cameras to cover the whole area close to the vehicle. Many
sensors are used for obstacle detection such as LIDAR, RADAR and vision sensors.
Pedestrian detection using moving camera
In present days, Sudden Pedestrian Crossing (SPC) is one of the major reason for
accidents on roads. Jegham and Khalifa have asserted that to reduce the death on road should
develop an Intelligent Transportation System (ITS). So that system will boost up the field of
computer vision. Mainly ITS system includes Advanced Driver-Assistance System (ADAS).
The system plays a major role and make sure that vehicle, pedestrian safety, drivers, and
passengers will be more comfortable. In the current paper, the author currently conducted
research on pedestrian detection under the weather condition such as foggy, snowy, and rainy
day. The author proposed one of the approaches to detect pedestrian [1]. From the proposed
research, the results are identified that approach provides great performance.
Pedestrians and Vehicles Recognition Based on Laser Distance
Nowadays, most of the vehicles are armed with a dashboard camera and display
devices. According to the survey, it is identified that accidents are increasing. Lin and Lee
stated that to recognize pedestrians and vehicle most of the researchers used a laser range
finder. It mainly finds the information of azimuth angle and also the distance of an object.
The author used a different kind of methods such as coordinate matching, breakpoint
detection, and hears detection. From the results, it is identified that the system is more
accurate and decrease errors [2]. In the article, the author provided clear quality information
and it is appropriate to the chosen topic.
Human Detection Robot using PIR Sensors
Human detection robot will detect the presence of a human. Mainly robot will move
in all the directions to improve space of detection. Sravana, Priscilla, Jose, and Balagoapal
asserted that using the PIR sensor can detect the human. Robot mainly consists of two sides
such as the transmitter side and receiver side. To move from left to right attached with a DC
motor. Human detection robot will provide security for many of the users in order to protect
their belongings. If the robot is attached with "Sound Navigation and Ranging" then it defines
distance among human [3]. In the current research paper, the author clearly stated about the
robot and provided journals and conferences.

7
RESEARCH METHODS AND PROJECT DESIGN
Pedestrian Behavior in Urban Scenarios
The ability to forestall pedestrian actions on the roads is an issue for the cars and it
has been increased in the automotive industry. According to Ridel, Rehder, Lauer, stiller, and
Wolf, mainly estimation of pedestrians while crossing roads is a challenging task. From the
study, several researchers had focused on the development of a system which enhances the
safety of pedestrians on roads. The dataset has been created for tracking and detection of
pedestrians in the image. Intelligent car application mostly requires precision and very high
accuracy [4]. From the paper, it is understood that the author had provided appropriate
information which is related to the chosen topic.
A Multi-Sensor Fusion System for Detection and Tracking pedestrians
The self-driving car needs to be deployed in a real world that must be capable of
effective tracking and detecting the moving objects. The author asserted that detection and
tracking of the object is a core task in the robotics and also in the field of intelligent vehicle.
Mainly to evaluate the performance of a sensor tracking system collected data such as images
and LIDAR scans. According to the results, it identified that tracking system shows an
effective performance on complete data [5]. In order to enhance the earlier system,
redesigned a sensor configuration and installed it into radar.
Pedestrian Detection System in order to Avoid Pedestrian Vehicle
Pedestrian detection is the most important field in both government and commercial
organizations. The author stated that night is possible with the combination of two different
approaches such as one is sufficient intensity range and other spectral range. Pedestrian
detection and tracking system is developed by simple segmentation in order to decrease the
vehicle-related accidents. In this author had used MATLAB tool to stimulate the system. But
the author stated that detection is a more difficult challenge in object processing field.
Majorly segmentation consists of step edge detection which ci algorithm applied to process
data regarding the shape of the pedestrian [6]. Finally, pedestrian detection is to reduce
pedestrian vehicle-related accidents. In the article, author clear define about the related topic
by using article and conferences. Completely explained about tracking and detection of a
pedestrian on images.
Promise and limits exploration regarding the pedestrian detection of automated
vehicles and pedestrian safety