International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-8 Issue-3, September 2019
7895
Published By:
Blue Eyes Intelligence Engineering
& Sciences Publication
Retrieval Number: C6564098319/2019©BEIESP
DOI:10.35940/ijrte.C6564.098319
Abstract: Video surveillance data in smart cities needs to analyze a
large amount of video footage in order to locate the people who
are violating the traffic rules. The fact is that it is very easy for the
human being to recognize different objects in images and videos.
For a computer program this is quite a difficult task. Hence there
is a need for visual big data analytics which involves processing
and analyzing large scale visual data such as images or videos.
One major application of trajectory object detection is the
Intelligent Transport Systems (ITS). Vehicle type detection,
tracking and classification play an important role in ITS. In order
to analyze huge amount of video footage deep learning algorithms
have been deployed. The main phase of vehicle type detection
includes annotating the data, training the model and validating
the model. The problems and challenges in identifying or
detecting type of vehicle are due to weather, shadows, blurring
effect, light condition and quality of the data. In this paper deep
learning algorithms such as Faster R CNN and Mask R CNN and
Frameworks like YOLO were used for the object detection.
Dataset (different types of vehicle pictures in video format) were
collected both from in-house premises as well as from the Internet
to detect and recognize the type of vehicles which are common in
traffic systems. The experimental results show that among the
three approaches used the Mask R CNN algorithm is found to be
more efficient and accurate in vehicle type detection.
Keywords : Deep learning, Intelligent Transport Systems, Mask
RCNN, YOLO
I. INTRODUCTION
Image processing is the process of analyzing and
manipulating an image by using a computer algorithms
in-order to extract useful information or to enhance the quality
of the image. This can be used in the process of finding
instances of the real world objects such as human, animals,
birds, faces, bicycles, car, truck and building etc. Video
content analysis (VCA) paves way to analyze and detect the
various temporal and spatial details available in the video.
Instead of analysis the still image, video contents can be
analyzed in-order to extract useful insight from it. Insights
from VCA can be used in various domains such as health care,
home automation and automotive industry in-order to provide
safety and security. Trajectory Object detection and
classification is a part of VCA. There is an urgent need for
intelligent transport systems to replace human operation to
monitor the surveillance area and to analyze a large amount of
video footage in order to locate the people who are violating
Revised Manuscript Received on September 19, 2019.
S Anitha Elavarasi*, Department of Computer Science & Engineering,
Sona College of Technology, Salem, India. Email:
anithaelavarasi@sonatech.ac.in
J Jayanthi, Department of Computer Science & Engineering, Sona
College of Technology, Salem, India . Email: jayanthij@sonatech.ac.in
N Basker, Department of Computer Science & Engineering, Sona
College of Technology, Salem, India . Email: baskern@sonatech.ac.in
the traffic rules. Huge volume of video footage has to be
analyzed in traffic control system especially in smart city
application. There is a need for deep learning algorithm to
easy the task. Deep learning is a part of machine
learning methods based on artificial neural networks. Deep
Learning algorithm such as Convolutional Neural Network
(CNN)[9], Faster R-CNN [6] has been widely used in the field
of object detection, segmentation and classification. The
CNN approach consists of neurons, activation function and an
output. Neuron contains learning units. The neurons in CNN
receive several inputs, take a weighted sum over them, pass it
through an activation function and finally respond with an
output. The most common challenges associated with
detection of vehicles are due to weather condition, shadows,
blurring effect, light condition, type of vehicle and quality of
the input data etc.
In this paper section 2 describes the related work and
section 3 explains the architecture of Trajectory Object
detection and classification system. Section 4 describes the
results and discussion and finally section 5 concludes the
work.
II. RELATED WORK
Yong Tang et al present vehicle detection and recognition.
The author uses automatic monitoring digital cameras to take
snapshots of moving motion pictures [1]. From the collected
images, sequences are extracted to represent features of a
vehicle. Initially machine learning algorithms like Haar-like
feature and Adaboost algorithm are applied for feature
extracting and constructing classifiers which is used to locate
the vehicle over the input image. The drawback of the system
is that it only works for day light image.
Viktoria Plemakova et al describe vehicle detection based on
convolutional neural networks [3]. The main aim of the author
is to train, classify and detect the vehicles from different
angles using neural networks. The proposed CNN contains 6
convolutional layers and 5 max pooling layers. The
challenges faced by the author are due to weather, light
conditions and vehicle type diversity.
Yilmaz describes the vehicle detection using deep learning
methods. The method takes five stages such as (1) loading the
data, (2) the design of the Convolutional Neural Network, (3)
training and configuration, (4) training of the R-CNN
approach and (5) evaluation of the detector [4]. The
parameters to be considered are image windows, selective
search for object recognition, category independent object
proposals, object segmentation using constrained parametric
min-cuts and multiscale
combinatorial grouping.
Trajectory Object Detection using Deep
Learning Algorithms
S Anitha Elavarasi, J Jayanthi, N Basker