Classification and Detection of Vehicles using Deep Learning

International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 4 Issue 3, April 2020 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
@ IJTSRD | Unique Paper ID – IJTSRD30353 | Volume – 4 | Issue – 3 | March-April 2020 Page 283
Classification & Detection of Vehicles using Deep Learning
Madde Pavan Kumar, Dr. K. Manivel, N. Jayanthi
Department of Electronics and Communication, Mahendra Engineering College, Namakkal, Tamil Nadu, India
ABSTRACT
The vehicle classification and detecting its license plate are importanttasksin
intelligent security and transportation systems. The traditional methods of
vehicle classification and detection arehighlycomplexwhichprovidescoarse-
grained results due to suffering from limited viewpoints. Because of thelatest
achievements of Deep Learning, it was successfully applied to image
classification and detection of objects. This paper presents a methodbasedon
a convolutional neural network, which consists of two steps: vehicle
classification and vehicle license plate recognition. Several typical neural
network modules have been applied in training and testing the vehicle
Classification and detection of licenseplatemodel,suchasCNN (convolutional
neural networks), TensorFlow, Tesseract-OCR. The proposed method can
identify the vehicle type, number plate and other information accurately.This
model provides security and log details regarding vehicles by using AI
Surveillance. It guides the surveillanceoperatorsandassistshumanresources.
With the help of the original (training) dataset and enriched (testing) dataset,
the algorithm can obtain results with an average accuracy of about 97.32% in
the classification and detection of vehicles. By increasing the amount of the
data, the mean error and misclassification rate gradually decreases. So, this
algorithm which is based on Deep Learning has good superiority and
adaptability. When compared to the leading methods inthechallengingImage
datasets, our deep learning approach obtains highly competitive results.
Finally, this paper proposes modern methods for the improvement of the
algorithm and prospects the development direction of deep learning in the
field of machine learning and artificial intelligence.
KEYWORDS: Convolution neural network, Vehicle classification, Vehicle License
plate Recognition
How to cite this paper: Madde Pavan
Kumar | Dr. K. Manivel | N. Jayanthi
"Classification & Detection of Vehicles
using Deep Learning" Published in
International Journal
of Trend in Scientific
Research and
Development
(ijtsrd), ISSN: 2456-
6470, Volume-4 |
Issue-3, April 2020,
pp.283-291, URL:
www.ijtsrd.com/papers/ijtsrd30353.pdf
Copyright © 2020 by author(s) and
International Journal ofTrendinScientific
Research and Development Journal. This
is an Open Access article distributed
under the terms of
the Creative
CommonsAttribution
License (CC BY 4.0)
(http://creativecommons.org/licenses/by
/4.0)
I. INTRODUCTION
Vehicle classification and detection have promising
importance in the future. Because, it can be utilized in many
aspects such as to analyze the urban traﬃc statistics,
Advanced Driver Assistance System, against vehicle theft,
large parking lot management, against vehicle escape,
improvement of the road information acquisition and safety
management of highway, electronic toll collection (ETC),
traﬃc investigation and so on.
One of the most important applications for object
recognition in the neural network is the image classification
tasks. However, the vehicle classification and detecting its
license plate still face a few great challenges because the
number of vehicle classes are very large and that some
attributions of the vehicle are too close to identify. So, the
process of training the neural network requires a great
number of parameters, which often accompanies by over-
fitting problems. The recognition results can achieve a
higher degree of accuracy in the training set, but a lower
degree of accuracy in the test set.
Fig 1: Classification of Vehicle
IJTSRD30353

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
Although the training data set is modest in size, the neural
network itself does not embody distinct advantages in
evaluation systems of higher recognition accuracy and
shorter training time than other models. The predominant
motives for it are limited computing sources and lengthy
processing time for even a small network. However, one of
the keys to these issues is the feature extraction and
classification of detection images for vehicle rearview.
Because of Facing many practical problems, neural networks
have been gradually integrated with some better methods
emerging in the 21st century, such as Convolutional Neural
Networks, ResNet Architecture, Tensor Flow. In this kind of
model, aimed at different tasks,differentsystemsareusually
designed and respectively applied different manual design
features.
Fig 2: License Plate Recognition
For example, the object recognition uses Feature extraction
from CNN. License platerecognitionusesTesseract-OCR,and
the pedestrian detection uses Histogram of Oriented
Gradient (HOG) feature. Therefore, this paper aims at
classifying and detecting the vehicles from different
viewpoints under any circumstances.
II. EXISTING SYSTEMS
The traditional methods of vehicle classification and
detection are mainly based on the following methods: 1)
SIFT (Scale-invariant feature transform) feature matching
and extraction; 2) the moving vehicle detection method of
Gaussian mixture model; 3) the license plate classification
method; 4) the monitoring video classification method of
HOG (Histogram of Oriented Gradient) and SVM (Support
Vector Machine). Based on these traditional methods, the
neural network model gets the lower recognition reference
to detect the objects. So, these methods give the inaccurate
results in which accuracy is limited to 85% only.
The above traditional algorithmshavealreadyobtainedgood
results. Several shortcomings in the traditional methods
have limited its realization: (i) the high similarity between
models naturally influence the accuracy, (ii) some models
containing only a few images, and (iii) the unified direction
like front, side or back in picture causes the inaccurate
classification. (iv) the methods usually require the images
are captured from certain viewpoints, (v) the vehicle
recognition granularity just stays in the stage of the vehicle
model classification. In this paper we address these
limitations and providing techniques that are practical for
vehicle classification problem. According to existing
methods, this paper further optimized the classification and
detection algorithms. It applies the Deep Neural Networks
(DNNs) to the classification and License plate detection of
the vehicles and obtained a better recognition performance
under different viewpoints or traffic conditions.
III. LITERATURE SURVEY
[1] Chen, Z., Ellis, T., and Velastin, S. A., “Vehicle Type
Categorization” A comparison of classification schemes In
Intelligent transportation systems(ITSC),14thinternational
IEEE conference.
This Paper proposed the Vehicle detection and classification
based on histogram oforientationgradients(HOG)approach
which was presented by Zezhi Chen and Tim Ellis. This
Classification model contains two parts: feature extraction
and classifier selection. In their approach, measurement-
based features (MBP) and the histogram of orientation
gradients (HOG) features are used to classify the vehicles
into four categories: car, van, bus, and motorcycle.
For this classification model, they have used two classifiers.
They are Random Forest (RF) and Support Vector Machines
(SVM). Due to the multiple classification problems, they
decided to use one-vs-all strategy. With this approach, the
best 3D model can be found that match with the original
vehicle type. The model they have presented is a novel
vehicle type classification, which provides an accurate and
reliable performance. However, the features among thefour
vehicle types are verydifferent.Sometimes3Dmodelsoftwo
vehicles such as hatchback and SUV may be similar. In this
case, this system may be hard to detect the vehicle type.
[2] Chang, S.-L., Chen, L.-S., Chung, Y.-C., and Chen, S.-W.,
“Automatic license plate recognition” IEEE transactions on
Intelligent Transportation Systems.
This paper proposed a license plate image technique
consisting of two main models:a licenseplatelocatingmodel
and a license number identification module. Specifically, the
license plates extracted from the first model areexamined in
the identification model to reduce the error rate. They used
color edge detection to compute edge maps.
But this is limited to only four types of edges. By using the
unique formulas, the model can transform RGB space into
HSI space that denote red, green, blue colors as hue,
saturation, intensity parameter values of an image pixels,
respectively. The identification module consists oftwomain
stages, preprocessing and recognition. After this process,
segmentation and recognition will be invoked sequentially.
However, this identification model takes more time to
recognize the characters and it is a complex process which
needs to be modified. This model can detect the License
plate, if and only if that license plate & characters are in
specified color edges. This type of model can be useful in a
particular region or place.
[3] Farhat, A., Al-Zawqari, A., Al-Qahtani, A., Hommos, O.,
Bensaali, F., Amira, A., and Zhai, X., “Tesseract-OCR Based
Feature Extraction and Template Matching Algorithms For
Number Plate Recognition” In Industrial Informatics and
Computer Systems (CIICS), International Conference on
IEEE.
A modified template matching Correlation algorithm was
proposed by Ali Farhat et al. They have developed four

algorithms for Numeric AutomaticLicensePlateRecognition
(ALPR) systems: Vector crossing, zoning and combined
zoning-vector and template matchingcorrelation.Firstthree
algorithms are based on feature extraction techniques and
the last one is application of correlation technique. By using
vector crossing algorithm, they distinguished the ten
characters (0−9) except the characters “2”, “3” and“5”.Since
these characters have the same number of the vectors. By
using Zoning method, densities of the image ineachzoneare
derived by the algorithm to determine the characters.
The third approach is the combination of the previous two
methods. In that, density oftheimageiscalculated.However,
these three algorithms cannot determine a noise in the
image. Hence, they developed a system based on template
matching correlation as a fourth approach. This correlation
indicates the linear relationship between each character
with all other characters. However, the limitations of this
approach are that, they can recognize only 10 numbers
instead of the alphabets. If more templates are added to the
system, the success rate may decline since some characters
share the same density and vector.
[4] Casey, R. G. and, E., “A survey of methods and strategies
in character segmentation”, IEEE transactions on pattern
evaluation and machine intelligence.
Lecolinet et al proposed a new algorithm for license plate
identiﬁcation. For faster detection of license plate regions,
they developed a novel method called Sliding Concentric
Windows (SCW), which can describe the irregularity in the
image based on image statistics. The SCW segmentation
algorithm involves following steps. (i) Two concentric
windows A and B were created fortheupperleftcornerpixel
of the image. (ii) Mean values are obtained from both
windows. In this approach, the RoI (Region of Interest) will
be detected if the ratio of the statistical measurements (B to
A) in the couple of windows reaches a threshold value.
Image masking, Binarization and connected components
analysis (CCA) functions are employed in sequence. As a
result, license plate region is detected. In license plate
processing stage, they created an approach to segment the
characters. The probabilistic neural network (PNN) is
invoked to recognize the characters. As a result, possible
character is determined in the output layer.
However, the proposed system relies highly on the lighting
condition and the physical appearance of the plates. Insome
cases, the model cannot detect the plate due to the
illumination condition or the quality of the plates.
IV. PROPOSED SYSTEM
This paper proposed an Advanced vehicle classification and
detection method based on deep neural networks. In this
system, vehicle classification task is accomplished by
Convolutional Neural Networks.Itextractsthevehicleregion
from its background area by pictures, in order to test the
generalization ability about the domestic vehicle types of
vehicle dataset. So, the vehicle region is segmented to
remove background noise to improve the recognition
accuracy. In this approach, theconvolutional neural network
is used to complete feature extraction and classification
training of labeled image datasetsofmanyvehicletypessuch
as bus, car, bike, bicycle. In this way, the task of identifying
vehicle types from different angles can be completed.
Fig 3: Input Image to the Model
Besides, this paper alsofocusesonautomaticdetectionofthe
vehicle License plates which can be accomplished by
Tesseract-OCR (optical character recognition). It is used to
recognize the characters and provides the text from the
given image. So, the extracted License Plate is given to the
tesseract-OCR which in return provides the characters of
number plate for license plate recognition.
Fig 4: Output Image from Trained Model
Finally, the Proposed model is evaluated based on selection
accuracy, recognition accuracyandreal-timeperformancein
the vehicle classification and detection process. While
training the model for the detection of vehicle type and its
number plate, the dataset must be split into train data and
test data. The train and test data are statistically known as
positive overlap range and negative overlap range
respectively.
When we dataset the parameter "Negative Overlap Range"
(where the overlap ratio between the candidate and the
image labeled area is used as the interval for negative
training samples) to [0, 0.3], and the parameter "Positive
Overlap Range" (the overlap ratio between the candidate
and the image labeled area as the interval of the positive
training sample) to [0.7, 1.0], the accuracy of proposed
vehicle detection model will be the highest.
V. SYSTEM ARCHITECTURE
Block Diagram of the Proposed System is represented as
shown below in figure 5. This Proposed block diagram is
categorized into threeparts. Theyare(i)SystemFramework,
(ii) Software Requirements, (iii) Methodology

Fig 5: Block Diagram of Proposed System
A. SYSTEM FRAMEWORK
1. Vehicle Detection and Classification Based on
Convolutional Neural Networks:
1.1. Vehicle Detection:
In this system, images of the training dataset are given as
input and the foreground area of the target vehicle is
detected by the convolutional neural network (CNN). But
there are some issues to build a deep learning model by
using CNN. The first problem while training the deeper
networks is, accuracy should be increased with an increase
in depth of the network as long as over-fitting is taken care
of. However, the problem with increased depth is that the
signal required to change the weights, which arises fromthe
end of the network by comparing ground-truth and
prediction becomes very small at the earlier layers, as a
result of increased depth.
It essentially means that earlier layers are almost negligible
to learn. This is called a vanishing gradient. The other
problem while trainingthedeepernetworksis,executingthe
optimization on huge parameter spaceandtherefore naively
adding the layers which results in higher trainingerror.This
is called a degradation problem.Inordertoovercomethese
issues, Convolutional Neural Network uses ResNet
Architecture to train the deeper networks.
ResNet architecture makes use of shortcut connections to
solve the vanishing gradient and degradation problems. The
basic building block of ResNet is a Residual block which is
repeated throughout the network. The ResNet structure as
shown in Figure 6.
Fig 6: A Residual Block of Deep Residual Network
Instead of learning the mapping from x → F(x), the network
learns the mapping from x→ F(x)+G(x).Whenthedimension
of the input x and output F(x) is same, the functionG(x)=x is
an identity function and the shortcut connection is called
Identity connection. The identical mapping is learned by
zeroing out the weights in the intermediate layer during
training, since it’s easier to zero out the weights than push
them to one. For the case when the dimensions of F(x) differ
from x, Projection connection is implemented rather than
Identity connection. The function G(x) changes the
dimensions of input x to that of output F(x).
Fig 7: Projection connection of Residual Network
ResNet Architecture uses Rectified Linear Unit (ReLu) as an
activation function of neural networks. This rectified Linear
activation function is a piecewise linear function that gives
the output as its input itself, if the input is positive.
Otherwise, it gives output as zero if the input is negative.
ReLu activation function reduces the large number of
computations. It has become the default activation function
for many types of neural networksbecausea model that uses
ReLu is easier to train and often achieves better
performance. The ReLu activationfunctionisgivenasshown
below.
X if x > 0
F (x) = for all x ∈∈∈∈R
0 if x 0
By using ReLU as an excitation function instead of the
traditional function, Logistics makes it easier to forward
propagation and calculate the reverse gradient by using
partial derivative. It also helps avoid complex algorithms
such as exponent and division.
The hidden layer neurons with output less than 0 are
discarded to increase the sparsity and reduces the
overfitting effect. The ResNet network structure is adjusted
while testing the data and the extracted vehicle area is
outputted as shown in Figure 8.

Fig 8: Vehicle Detection
1.2. Vehicle Classification:
The foreground vehicle area obtained by vehicle detection
algorithm is introduced into another convolutional neural
network for vehicle recognition. Convolutional neural
network uses a certain number of convolutional kernels to
slide the input image with a certain step size to extract
features in the convolution layer. Each convolution kernel
focuses on different features, so the convolution obtains
different feature maps.
Fig 9: CNN Architecture
These obtained feature maps are passed into the lower
convolution layer, and then the lower convolutionlayeruses
a certain number of convolution kernels to extract the
features of upper feature map, this operation will be
repeated. The weighting parameter will be got after the
features are processed for several times. Thentheweighting
parameter will be connected to the full connection layer, the
full connection layer plays a role of classifier in the entire
convolution neural network, namely a model to classify the
extracted features.
By using the method of Deep Learning, the features ignored
by the traditional methods can be extracted, the recognition
accuracy can be improved obviously, and different
convolution kernels are used to obtain convolution feature
diagrams, as shown in Fig 10. This ResNet architecture is
similar to the VGGNet which consists mostly of 3X3 filters or
kernels.
Fig 10: Feature Map of Convolution Layer
From the VGGNet, shortcut connection is inserted to form a
residual network. In order to overcome the issues such as
vanishing gradient and degradation problems of the deep
learning model which occurs during training the model.
2. TENSORFLOW:
TensorFlow is an open source library which uses data flow
graphs to build the models. This allows developers to create
large-scale neural networkswithmanylayers. TensorFlowis
a neural network module which is mainly employed for
Classification, Perception, Understanding, Discovering,
Prediction and Creation. Tensors are the multidimensional
arrays, an extension of 2-dimensional tables to data with a
higher dimension. There are many features regarding
TensorFlow which makes it appropriate for Deep Learning.
Fig 11: TensorFlow Architecture
TensorFlow architecture works in three parts. They are (i)
Preprocessing the data, (ii) Build themodel,(iii)Totrainand
test the model. Initially, the raw data is scaled by
normalizing the input and output features of a model. This
preprocessed data splits into train and test dataset. This
Building model is trained using train dataset. By comparing
the prediction accuracyoftestdatasetwiththeground-truth,
loss function of the model can be analyzed and tuned with
hyperparameters. TensorFlow is used for fast numerical
computations of neural networks. It is a foundation library
that can be used to provide DeepLearningmodelsdirectly or
perhaps by using wrapper libraries that simplifytheprocess
built on top of TensorFlow.
3. OpenCV:
OpenCV is also known as Open Source Computer vision. It
is a cross-platform library of programming functions
which mainly aims to develop the real-time computer
vision applications. OpenCV was developed by Intel
organization. This mainly focuses on image processing,
video capture and analysis which include features such as
face detection and object detection. It supports some
models from deep learning frameworks such as tensor-flow,
Tesseract-OCR, PyTorchmodulestobuildaneffectivemodel.
By integrating the OpenCV with the Tesseract-OCR, it can
extract the characters even from colored imageswithhigher
accuracy.
B. SOFTWARE REQUIREMENTS
1. TESSERACT-OCR:
The Character Recognition popularly referred as Optical
Character Recognition (OCR). Tesseract-OCR is the
software process of converting typed, handwritten, or
printed text to machine-encoded text that we can access
and manipulate via a string variable. The OCR has been
one of the challenging and popular ﬁelds of research in
pattern recognition. This can be used as one of the

classifier in the image processing step since the most
crucial step in OCR is identifying characters as one of 36
symbols (A-Z and 0-9). Therefore, for classification, we
need to extract visual features from individual character
image. Then, character can be classified with these
features using machine learning. The most effective way
for now is OCR-tesseract, which is originally developed by
Hewlett Packard in the 1980s. It is one of the most
accurate open source OCR engines currently available.
Fig 12: Flowchart of Tesseract-OCR Process
As other traditional procedures, this processing follow a
step by step approach. As the most important part,
recognition consists of two stages, which are adaptive
classification and repeating recognition, respectively. In
the first stage, each character is recognized sequentially.
Additionally, the symbol that is classified is stored by the
adaptive classifier as the training data. Then, the adaptive
classifier will learn the information that classified
characters provide, from which the characters that are
not recognized in first step will be classified again. This
two-stage process make this method accurate and
efficient.
2. LabelIMG:
Labelimg is a graphical image annotation tool. It is designed
in Python and uses Qt for its graphical interface.Annotations
are saved as XML files in PASCAL VOC format which is
employed by ImageNet. Image Annotation could be
the process of building datasets for computervisionmodels.
This helps machines to learn, how to automatically assign
metadata into a digital image using captioning or keywords.
This Labelimg tool is used to train our customized model.
This technique is used for image retrieval systems to
organize and easilylocateparticular images froma database.
Image labeling gives the insight into the content of images.
When you use the API, you get a list of the entities that were
recognized: people, things, places, activities, and so on.
Each label found comes with a score that indicates the
confidence or probability or accuracy of the ML model
according to its relevance.
3. ANACONDA NAVIGATOR:
Anaconda Navigator is used to launch applications and to
manage conda packages, virtual environments, and
programs without the use of command line commands. In
order to get the Navigator, download the Navigator Cheat
Sheet and install Anaconda. It is a free open-source software
which is associated with Python and R programming
languages for scientific computing such as for data science,
machine learning applications, large scale data processing,
predictive analytics and so on. It aims to simplify package
management and deployment.
Jupyter Notebook is a one of the IDE which have to be
launched from the Anaconda Navigator to develop, execute,
debug the algorithms which performs Artificial intelligence
tasks. It provides you with an easy-to-use, interactive data
science environment across many programming languages
that doesn't only work as an IDE, but also as a presentation
or education tool. It's perfect for those who are just starting
out with data science. Because, it helps to debug the errors
by specifying the type of error that was obtained.
C. METHODOLOGY
1. DATA COLLECTION & DATA PREPARATION:
In this Project, dataset is prepared by the collection of
vehicle images from Open source, CCTV cameras and
internet. To ensure the representativeness and the integrity
of data, images are taken from different viewpoints under
different traffic conditions. Most of theimagesinthisdataset
are rear views of the vehicle. Besides, the image dataset also
covers most types of vehicles and noise images.
This involves collection of data by recording using a camera
with auto iris function which keeps the average illumination
of the view constant. It also uses i-LIDS dataset, then the
input data is processed into a set of features before
becoming suitable inputs for perframevehicledetectionand
classification using 3D models.
2. EDGE DETECTION:
An edge based multi-stage detection is the main primary
function to detect the vehicle and its license plate edges at
their localization. In this paper, license plate edges are
detected from its original image by edge detection.
Localization of plate is undoubtedly a challenging task
since there are significant variations in plate size, color,
lighting condition and spatial orientationsof license plate
in images. So, three steps are employed at the
preprocessing stage. They are (i) Gray Scale Conversion,
(ii) Median Filtering, (iii) Contrast Enhancement.
2.1. Gray Scale Conversion:
By using following formula, the 24-bit color image can be
converted into 8-bit gray image.
Gray = 0.59×R + 0.30×G + 0.11×B
2.2. Median Filtering:
As one of non-linear filter, it can calculate the median of
the gray values of a pixel’s neighbors. In this stage, they
use 3×3 masks to get eight surrounding neighbors’ gray
value and replace the pixel value with the median value.
As a result, the function could remove salt-and-peeper
noise from the image.

2.3. Contrast Enhancement:
Histogram equalization technique is invoked to enhance
contrast of the images. In the procedure of the conversion,
the total number of pixels in the image is N and the
number of pixels with the gray level k is nk. Then the
stretched gray level Sk is calculated by the following
formula.
Sk =
3. FUNCTIONAL PROCESS:
This paper proposed an Advanced vehicle classification and
detection method based on the deep learning approach. In
this system, the surveillance cameras are used for observing
the vehicles at any specific region orarea.Thesesurveillance
cameras are often connected to the Personal Computer (PC)
or any recording device to provide those vehicle images or
videos which are passing through that specific gateway or
area. This vehicle data is taken as input to the Personal
Computer (PC). By using the i-Lids, real time camera and
open source data a dataset is created. With the help of this
dataset, the Deep Learning model istrainedup-toa specified
epoch. With the Increase of number of epochs, the loss
function of the model is reduced.Finally,thetrainedmodel is
tested with the new vehicle data to check whetherthemodel
is working effectively in vehicle classification and detection
of their number plates.
Fig 13: Functional Process of Proposed Scheme
In this system, Vehicle classification method is based on a
convolutional neural networks and automatic license plate
detection of vehicles method is based on Tesseract-ocr. This
Deep neural network model will detect the vehicle from the
given image or video by using frame extraction. It also
provides the classified type of vehicle whether it is bus or
bike or bicycle or any other type.
The model will detect the vehicle number plate if and only if
there is number plate for that detected vehicle. Thedetected
vehicle number plate is extracted as an imagefromthegiven
data and simultaneously giventotheTesseract-OCR whichis
integrated with the OpenCV to recognize the characters
present in that number plate. The model will beinstructed to
detect and classify the vehicles, only when their detection
probability is more than 0.90. This Probabilityapproachwill
be useful to get the accurate vehicle info or effective results
of the model. A log is created in the PC, to note the details of
the vehicles.
This Info includes type of vehicle, its number plateandother
info like time in & time out are tabulated accurately in the
log. Every vehicle info is appended to its log. Finally, this
model provides the accurate info of log regarding the
vehicles that are passed through this Artificial Intelligence
Surveillance.
VI. FLOW CHART
Fig 14: Flow Chart of Proposed System
The given flow chart of proposed system explains the
procedure and sequence of Framework as shown above in
figure 14.
VII. EVALUATION & DISCUSSION
For evaluating the proposed or implemented model, we
must compare the performanceoftheproposedsystemtoall
other existing systems. After collecting the relevant training
and testing experiment information, we analyzed and
compared the accuracy rate of the different network
structures such as AlexNet, Vggnet16, Vggnet19, Googlenet,
Resnet50 and Resnet101. Currently, proposed system uses
ResNet101 architecture to train and test the deep learning
Algorithm for vehicle classification and detection. The
accuracy for different network architectures shown below.
CR (%) =
In the equation, CN means for the number of correctly
classified images and EN means for the number of
incorrectly classified images indicates the correct rate. To
compare the experimental resultsofvariousnetwork results,
the graphical representations of them are given below.

Fig 15: (a) loss curves of curves; (b) curves of accuracy
The dataset with 3500 pictures is collected and
preprocessed with auto-iris function which is used to
evaluate the trained models. We drew the loss curve and
accuracy curve of every classification and recognition
network structure, which are helpful to understand the
training process of every model. From the comparisonofthe
curves, it is clear that the Resnet-101 can get the fastest
convergence and higher accuracy as shown in Fig 15.
Architecture Accuracy Year
Alex Net 82.6 2012
Inception-V1 88.4 2013
VGG 90.8 2014
ResNet-50 92.3 2016
ResNet-101 95.2 2018
Table 1: Comparision of Network Structures
We apply the proposed model which is trained by the image
dataset into the real traffic road video or imagescollected by
ourselves. The collection of real traffic vehicle pictures is
also tested. We observed that for the recognition of the real
traffic road image, the training image dataset obtains the
higher accuracy.
The experimental results obtained from the training and
testing dataset is shown in the below table. The figure 15
shows that, Resnet Network structure gives the higher
accuracy for object detection and classification than the
other network structures. By analyzing the testing
performance with different number of layers in ResNet
structure, the best performing one is the cascading of the
ResNet101 network structure which achieves 95.27%.
VIII. RESULTS
In most of the cases, the experimental results of proposed
system have obtained with higher accuracy. The proposed
method has integrated feature extraction, object frame
boundary generation, linear regression & classification and
License plate recognition to provide a customized efficient
model. By using this model, we can retrieve the vehicle info
accurately. Even if there is background noise in the license
plate, it will not affect the performanceoftherecognition.So,
this system can recognize all the characters on the license
plate successfully.Thus,thecomprehensiveperformance has
been greatly improved. When compared to existing models,
the current deep learningvehicleclassificationanddetection
neural network model has the rate of accuracy as very high.
Therefore, the proposed model gives standard results.
IX. CONCLUSION
Based on CNN, this paper proposed the vehicle type
classification and license plate recognition in urban traffic
video surveillance. This Deep Learning model can do both
Classification and Detection of vehicles simultaneously.This
reduces the complexity of the processing which helps to
increase the performance of the system to train and test the
model. With the increaseoftrainingdataset,themodification
of parameters and replacement of the model; the proposed
method becomes effective and validated by the relevant
experiments. The experimental resultsshowthatthevehicle
recognition rate is improved. In a comparative analysis, this
vehicle classification & detection framework can get high
accuracy with success rates which has been proposed to
expose in a better performance than existing frameworks.
X. FUTURE WORK
In the future, we will implement this proposed model as a
real-time hardware application with a deep learning
Framework which further improves the accuracy and
robustness for the traffic vehicleclassificationanddetection.
There are still mountains of work about deep learning to be
studied. For example, it needs the other efficient and theory
based deep learning model algorithms. It explores the new
feature extraction models which are also worthy to do
further research. Besides studying efficient parallel training
algorithms, they need to be investigated. Regarding the
applications of extending deep learning and how to make
rational use of deep learning in enhancing the performance
of traditional algorithms is still the focusofa varietyoffields.
XI. ACKNOWLEDGEMENT
I would like to express my sincere appreciation to my
supervisors Dr. N. Viswanathan, Dr. K. Manivel and N.
Jayanthi for their constant guidance and encouragement
during my whole B.E period at the Mahendra Engineering
College. They gave me this opportunity to do this great
project on “CLASSIFICATION & DETECTION OF VEHICLES
USING DEEP LEARNING” which helps me to learn a lot of
new things. I am thankful to them; without their valuable
help, this project would not have been possible.
XII. REFERENCES
[1] Saha, S., Basu, S., Nasipuri, M., and Basu, D. K.,
“License plate localization from vehicle images: An
edge-based multi-stage approach”, International
Journal of Recent Trends in Engineering, 1(1):284–
288.

[2] Chang, S.-L., Chen, L.-S., Chung, Y.-C., and Chen,S.-W.,
“Automatic license plate recognition”, IEEE
transactions on intelligent transportation systems,
5(1):42–53.
[3] Al-Shami, S., El-Zaart, A., Zantout, R., Zekri, A., and
Almustafa, K. “A new feature extraction method for
automatic license plate recognition”, In Digital
Information and Communication Technology and its
Applications (DICTAP), Fifth International IEEE
Conference.
[4] Farhat, A., Al-Zawqari, A., Al-Qahtani, A., Hommos, O.,
Bensaali, F., Amira, A., and Zhai, X., “Ocr based
feature extraction and template matching algorithms
for Qatari number plate”, In Industrial Informatics
and Computer Systems (CIICS), International IEEE
Conference.
[5] Anagnostopoulos, I. E., Loumos, C. N. E.,
Anagnostopoulos, V., and Kayafas, E., “A license plate
detection algorithm for intelligent transportation
system applications”, IEEE Transactions on
Intelligent transportation systems, 7(3):377–392.
[6] Dalal, N. and Triggs, B., “Histograms of oriented
gradients for human detection In Computer Vision
and Pattern Recognition”, CVPR IEEE Computer
Society Conference on, volume 1, IEEE.

Classification and Detection of Vehicles using Deep Learning

Recommended

More Related Content

What's hot (20)

Similar to Classification and Detection of Vehicles using Deep Learning (20)

More from ijtsrd (20)

Recently uploaded (20)

Classification and Detection of Vehicles using Deep Learning