AN INNOVATIVE RESEARCH FRAMEWORK ON INTELLIGENT TEXT DATA CLASSIFICATION SYSTEM USING GENETIC ALGORITHM

International Journal of Artificial Intelligence and Applications (IJAIA), Vol. 7, No. 6, November 2016
DOI: 10.5121/ijaia.2016.7604 43
CENTROG FEATURE TECHNIQUE FOR VEHICLE
TYPE RECOGNITION AT DAY AND NIGHT TIMES
Martins E. Irhebhude, Philip O. Odion and Darius T. Chinyio
Faculty of Science, Department of Computer Science, Nigerian Defence Academy,
Kaduna, Nigeria.
ABSTRACT
This work proposes a feature-based technique to recognize vehicle types within day and night times.
Support vector machine (SVM) classifier is applied on image histogram and CENsus Transformed
histogRam Oriented Gradient (CENTROG) features in order to classify vehicle types during the day and
night. Thermal images were used for the night time experiments. Although thermal images suffer from low
image resolution, lack of colour and poor texture information, they offer the advantage of being unaffected
by high intensity light sources such as vehicle headlights which tend to render normal images unsuitable
for night time image capturing and subsequent analysis. Since contour is useful in shape based
categorisation and the most distinctive feature within thermal images, CENTROG is used to capture this
feature information and is used within the experiments. The experimental results so obtained were
compared with those obtained by employing the CENsus TRansformed hISTogram (CENTRIST).
Experimental results revealed that CENTROG offers better recognition accuracies for both day and night
times vehicle types recognition.
KEYWORDS
CENTROG, CENTRIST, Vehicle Type Recognition, Day-time Recognition, Night-time Recognition,
Classification
1. INTRODUCTION
A feature technique that can adequately recognise vehicle types at day and night times has been a
subject in most vision related researches. Most of the time, a feature technique that gave an
optimal recognition accuracy on day time experiment for vehicle type recognition often fails to
replicate the same accuracy result when the same feature is applied on same dataset in same
locality at night time. This is probably because, features that were used for the day time
experiment are mostly appearance related i.e. information about colour and texture appearance of
the vehicle; whereas a night time feature is mostly contour based.
As reported by [1], not much work have been done on vehicle recognition at night time on
thermal images. However, the authors proposed CENsus Transformed histogRam Oriented Gradient
(CENTROG) feature technique for vehicle recognition at night times. Also reported in [2], Iwasaki
et. al., a vehicle detection mechanism within thermal images using the Viola Jones detector was
proposed. The technique involved detecting the thermal energy reflection area of tires as a
feature. This paper therefore aims to contribute towards bridging this research gap by proposing a
technique that will recognise vehicles at both day and night times.

44
Also reported by [3] and documented by several other authors, vehicle at day time were detected
and recognised by using different approaches. A multiple feature techniques was proposed in [3]
for recognising vehicle types during the day time. The proposed feature was experimented on two
different view angles. A state-of-the-art video processing techniques for vehicle detection,
recognition and tracking was surveyed in [4]. A combination of salient geographical and shape
features of taillights and license plates extracted from the rear view of a vehicle was used in [5]
for the recognition of vehicle make and model. Extracting additional data from video stream,
besides the vehicle image itself helped improve vehicle recognition in a 3D image sets [6]. A
technique for road intersection classification was proposed in [7]. A comparative analysis
between Support Vector Machine (SVM) and deep neural network with sift features showed that
automatic feature extraction recorded higher accuracy compared with manual technique [8]. Li et
al in [9] showed that even in a congested road traffic condition an AND-OR graph (AOG) using
bottom-up inference can be used to represent and detect vehicle objects based on both front and
rear views. Similarly, [10] proposed the use of strong shadows as a feature to detect the presence
of vehicles in a congested environment. In a separate experiment, authors in [11] showed how
vehicles were partitioned into three parts; road, head and body, using a tripwire technique.
Subsequently Haar wavelet features extracted from each part with PCA performed on features
calculated to form 3 category PCA-subspaces. Further, Multiple Discriminate Analysis (MDA) is
performed on each PCA-subspace to extract features, which are subsequently trained to identify
vehicles using the Hidden Markov Model Expectation Maximisation (HMMEM) algorithm. In an
experiment by [12], a camera calibration tool was used on detected and track vehicle objects so as
to extract object parameters, which were then used for the classification of the vehicle into classes
of cars and non-cars. Vehicle objects were detected and counted using a frame differencing
technique with morphological operators: dilation and erosion [13]. In [14], the author added
simulated images to initial dataset and applied PCA on each sub-region to reduce feature sets,
computation time and hence speed-up the processing cycle. In another classification task in [15],
the foreground image size feature was extracted with two levels dilation and fill morphological
operations; and classified into small, medium and large categories. In [16], the author proposed
Scale-Invariant Feature Transform (SIFT), the Canny edge detector, k-means clustering with
Euclidean matching distance metric for inter and intra class vehicle classification as an alternative
to expensive Electronic Toll Collection (ETC) full-scale multi-lane free flow traffic system. In
[17], a technique for traffic estimation and vehicle classification using region features with a
neural network (NN) classifier was proposed. A technique for rear view based vehicle
classification was proposed in [18] with investigation of Hybrid Dynamic Bayesian Network
(HDBN) in vehicle classification. HDBN was compared with three other classifiers for the
classification of known vs unknown classes and four known classes of vehicles using tail light
and vehicle dimensions with respect to the dimensions of the license plate as feature sets.
Similarly, the width, distance from license plate and the angle between the tail light and the
license plates formed part of the eleven features sets.
As we can observe from literature, it is clear that different feature approach has been used for
recognising vehicle types at both day and night times respectively. Therefore, in this research
work, we will examine features that are contour and appearance based, so as to design a technique
that can recognise vehicle during the day and also at night times. This will help eliminate the
bottleneck associated with generating new feature sets for dataset at a given time including view
angle. Day time video dataset which contain vehicles of varied colours and shapes will be used
for the day time experiment while thermal video dataset will be explored for the night time
experiment.

45
The rest of the paper is organized as follows. Section 2 explains the foreground/background
segmentation algorithm, Gaussian Mixture Model (GMM), CENsus TRansformed hISTogram
(CENTRIST) and CENTROG descriptors along with their usage was discussed in section 3.
Section 4 provides an overview of the Support Vector Machine (SVM) classification algorithm.
An insight into the research approach is provided in section 5 and followed by experimental and
performance evaluation in Section 6. The paper is concluded in Section 7.
2. GAUSSIAN MIXTURE MODEL (GMM)
According to [19, 20], a GMM is a parametric probability density function that is represented as a
weighted sum of Gaussian distributions. The GMM technique uses a method to model each
background pixel by a mixture of k Gaussian distributions [21]. The weight of the mixture
represents the time proportion for which the pixel values stay unchanged in a scene. Probable
background colours stay longer and are more static than the foreground colours.
In [22], the recent history of each pixel, 1,..., tX X , is modelled by a mixture of K Gaussian
distributions. The probability of observing the current pixel value is defined as:
, ,
1 ,
( ) * ( , , )
K
t i t t i t
i i t
P X Xω η µ
=
= ∑ ∑ (1)
Where K is the number of distributions, ,i tω is an estimate of the weight (what portion of the data
is accounted for by this Gaussian) of the ith
Gaussian in the mixture at time t, ,i tµ is the mean
value of the ith
Gaussian in the mixture at time ,, i tt ∑ is the covariance matrix of the ith
Gaussian in the mixture at time t, and η is a Gaussian probability density function of the form:
1( )1
( )
2
1
2 2
1
( , , )
(2 ) | |
Xt tT
t tX
t n
X e
µ
µ
η µ
π
− −
∑
− −
=∑
∑
(2)
The covariance matrix is of the form:
2
,k t k
Iσ=∑ (3)
3. CENTRIST AND CENTROG DESCRIPTORS
As reported in [1] Census Transformed Histogram for Encoding Sign Information (CENTRIST)
is a visual description technique that was proposed by Wu et. al. [23] that is used to detect
topological sections or scene categories. It extracts the structural properties from within an image
while filtering out the textural details. It employs the Census Transform (CT) technique in which
an 8-bit value is computed in order to encode the signs of comparison between neighbouring
pixels. According to [24], CT is a non-parametric local transforms described as follows:

46
Let P be a pixel, I(P) its intensity (usually an 8-bit integer), and N(P) the set of pixels in some
square neighbourhood of diameter d surrounding P. All non-parametric transforms depend upon
the comparative intensities of P versus the pixels in the neighbourhood N(P).
Define ( , ')P Pξ to be 1 if ( ') ( )I P I P< and 0 otherwise.
Rτ(P) maps the local neighbourhood surrounding a pixel P to a bit representing the set of
neighbouring pixels whose intensity is less than that of P. Therefore, census transform compares
the intensity value of a pixel with its eight surrounding neighbours; in other words, CT is a
summary of local spatial structure given by equation (5) [24]:
Let N(P) = P, where ⊗ is the Minkowski sum and D is a set of displacements, and let ⊗ be
concatenation.
( ) ( , [ , ])
[ , ]
R P P P i j
i j D
τ ξ
⊗
= +
∈
(4)
Example:
2
1 0 0
1 1 (10011110) 158
1 1 0
CT⇒ ⇒ ⇒ = (5)
From the CT example above, it can be seen that if the pixel under consideration is larger than (or
equal to) one of its eight neighbours, a bit 1 is set in the corresponding location; else a bit 0 is set.
The eight bits generated from the intensity comparisons can be put together in the order of
appearance (from top to bottom, left to right) and converted to a base-10 value (e.g., binary to
decimal conversion). This is the computed CT value for the pixel under consideration. The so-
called CENTRIST descriptor therefore is the histogram of the CT image generated from an
image.
In [1], in order to compute the CENTROG features, after the image structure has been captured,
compute its CT on a computed edge image, thereafter Histogram Oriented Gradient (HOG) [25]
features is extracted from the transformed edge image. The HOG works by counting the
occurrences of gradient orientation in localized portions of an image. The HOG captures local
object appearances and shapes, which can often be characterized rather well by the distribution of
local intensity gradients, or edge directions as reported in [26]. The gradient is computed by
applying [1,0,1] and [1,0,1]T
in horizontal and vertical directions within an image. The gradient
information is collected from local cells and put into histograms using tri-linear interpolation. On
the overlapping blocks composed of neighbouring cells, normalisation is performed. The
CENTROG descriptor therefore is the HOG on the CT generated image. Some vehicle image

47
samples are shown below, (figures 1 and 2). CENTROG is a very useful technique which helps to
capture the local and global structure of a particular image effectively when the colour and
texture information are missing in it.
Figure 1: Samples Showing Night Time Vehicles
Figure 2: Samples Showing Day Time Vehicles
4. SUPPORT VECTOR MACHINE (SVM)
According to [27] SVM is a technique used to train classifiers, regressors and probability
densities that is well-founded in statistical learning theory. SVM can be used for binary and
multi-classification tasks.
4.1. BINARY CLASSIFICATION
SVM perform pattern recognition for two-class problems by determining the separating
hyperplane with maximum distance to the closest points of the training set. In this approach,
optimal classification of a separable two-class problem is achieved by maximising the width of
the margin between the two classes [28]. The margin is the distance between the discrimination
hyper-surface in n-dimensional feature space and the closest training patterns called support
vectors. If the data is not linearly separable in the input space, a non-linear transformation Φ(.)
can be applied, which maps the data points x ∈ R into a high dimensional space H, which is called
a feature space. The data is then separated as described above. The original support vector
machine classifier was designed for linear separation of two classes; however, to solve the
problem of separating more than two classes, the multi-class support vector machine was
developed.
4.2 MULTI-CLASS CLASSIFICATION
SVM was designed to solve binary classification problems. In real world classification problems
however, we can have more than two classes. In the attempt to solve q-class problems with
SVMs; training q SVMs was involved, each of which separates a single class from all remaining
classes, or training q2
machines, each of which separates a pair of classes. Multiclass
classification allows non-linearly separable classes by combining multiple 2−class classifiers.
N−class classification is accomplished by combining N, 2−class classifiers, each discriminating
between a specific class and the rest of the training set [28]. During the classification stage, a
pattern is assigned to the class with the largest positive distance between the classified pattern and

48
the individual separating hyperplane for the N binary classifiers. One of the two classes in such
multi-class sets of binary classification problems will contain a substantially smaller number of
patterns than the other class [28]. SVM classifier was chosen because of its popularity and speed
of processing. For more details on SVM algorithm, please refer to this article [29].
5. RESEARCH APPROACH
In order to recognise vehicles at both day and night times, an initial classification between day
and night time vehicles will be done. After this initial categorisation, further classification will be
done for the day and the night time datasets. Figure 3 shows the proposed methodology for
vehicle type recognition at day and night times. As shown in figure 3, the input video was
segmented using Gaussian Mixture Model (GMM) foreground/background segmentation
algorithm. This was done to enable the effective extraction of the region of interest which
contained vehicle objects which were the only moving objects in the video input.
Figure 3: Proposed System Outline for Vehicle Type Recognition at Day and Night Times

49
In the experiments, based on the number of vehicular samples available in the publicly available
night time dataset [30], we will only focus on two key types for night times: cars and trucks. For
the night images, we used a thermal camera with parameters FLIR SR-19 Thermal Camera,
White Box, Black Box, Total Video Footage Captured: 63 min of ROBB DRIVE and 1-80
OVERPASS. However, for the day vehicular images, data samples were gathered from a low
medium resolution camera that was installed on the roadside of the Sohar Highway, in Oman.
The camera was of pixel resolutions, 640 x 360, and the frame rate was 25 FPS. The data used in
the experimental analysis consisted of 10hours video footage during daytime at approximately a
450
angle from the direction of the movement of vehicles.
6. EXPERIMENTS AND PERFORMANCE EVALUATION
A number of experiments were conducted to evaluate the performance of the proposed algorithm
on vehicle type recognition at both day and night times. The experiments were conducted on
images retrieved from camera installed on a roadside in Sohar Highway, Oman was used for day
time vehicular experiments. Similarly, the video dataset given in [30] was used for night time
experiments. The results obtain from these experiments will be discussed in the following
subsections.
To evaluate the system approach, Receiver Operating Characteristic (ROC) curve will be used.
As reported in [3] ROC curve shows classification performance in detail. A ROC curve is the plot
of True Positive Rate and the False Positive Rate for different cut-off points or threshold of a
parameter [31]. It is given as:
True Positive Rate (Recall) =
( )
tp
tp fn+
(6)
False Positive Rate =
( )
fp
fn tn+
(7)
where, tp denotes the number of true positives (an instance that is positive and classified as
positive); tn denotes the number of true negatives (an instance that is negative and classified as
negative); f p denotes the number of false positives (an instance that is negative and classified as
positive) and fn denotes the number of false negatives (an instance that is positive and classified
as negative).
ROC curve visualises the following as reported:
1. It shows the trade off between sensitivity and specificity (any increase in sensitivity will
be accompanied by a decrease in specificity).
2. The closer the curve follows the left-hand border and then the top border of the ROC
space, the more accurate is the test.
3. The slope of the tangent line at a cutpoint gives the likelihood ratio (LR) for that value of
the test.

50
Similarly, the accuracy of an experiment is measured by the Area Under the ROC Curve (AUC).
An area of 1 represents a perfect test; an area of 0.5 or less represents a worthless test. Accuracy
of performance is defined as:
p n
p n p n
t t
Accuracy
t t f f
+
=
+ + +
(8)
The following is a rough guide for classifying the accuracy of a test is the traditional academic
point system as reported by [32]:
• 0.90-1 = excellent (A)
• 0.80-0.90 = good (B)
• 0.70-0.80 = fair (C)
• 0.60-0.70 = poor (D)
• 0.50-0.60 = fail (F)
Finally, the ROC curve shows the ability of the classifier to rank the positive instances relative to
the negative instances.
6.1. DAY/NIGHT VEHICULAR CATEGORISATION EXPERIMENTS
The set of input-output sample is
1 1 2 2( , ),( , ),...,( , )N Nx y x y x y (9)
where the input ix denotes the feature vector extracted from image i and the output iy is a class
label. Since we are categorising into day and night times vehicles, the class label iy encodes day
time vehicles (encoded as 1) and night time vehicles (encoded as 2) respectively; while the
extracted feature ix encodes image histogram. Image histogram feature was used for
categorisation because it effectively captures appearance information of any data. Also, there is a
clear distinction between day and night data.
The dataset consisted of approximately 1500 day and night time vehicles and was split 75:25 for
the purpose of training and testing. The vehicles captured and thus used in experimentation only
consisted of cars and trucks (night time vehicles) and cars, jeeps and trucks (day time vehicles)
hence the classification was of a binary nature, i.e. into these two classes.
Experiment results shows 100% recognition accuracy; this is due to the fact that there is
appearance distinction between the day and night time vehicles. The confusion matrix and ROC
curve of the experimental results are shown in figures 4 and 5 below.

51
Figure 4: Confusion Matrix for Day and Night Time Categorisation
Figure 5: ROC Curve for Day and Night Time Categorisation
6.2. Day Vehicular Categorisation Experiments
The set of input-output sample is
1 1 2 2( , ),( , ),...,( , )N Nx y x y x y (10)
where the input ix denotes the feature vector extracted from image i and the output iy is a class
label. Since we are categorising into vehicle types, the class label iy encodes cars (encoded as 1),
jeeps (encoded as 2), and truck (encoded as 3) respectively; while the extracted feature ix
encodes image CENTROG.

52
The dataset consisted of approximately 720-day time vehicles and was split 75:25 for the purpose
of training and testing. The vehicles captured and thus used in experimentation only consisted of
cars, jeeps and trucks hence the classification was of a multi-categorisation, i.e. into three classes.
Experiments were conducted using CENTROG and compared with CENTRIST feature
descriptor. Experimental results obtained showed that CENTROG (the proposed technique)
outperformed CENTRIST by recording a detection accuracy of 97.2% versus 95.6%. The
confusion matrix and ROC curve of the experimental results are shown below in figures 6, 7, 8
and 9 for the purpose of comparison.
Figure 6: CENTROG Confusion Matrix for Day Time Vehicle Type Recognition
Figure 7: CENTRIST Confusion Matrix for Day Time Vehicle Type Recognition

53
Figure 9: CENTRIST ROC Curve for Day Time Vehicle Type Recognition
Figure 8: CENTROG ROC Curve for Day Time Vehicle Type Recognition
6.3. NIGHT TIME VEHICULAR CATEGORISATION EXPERIMENTS
In the experiments conducted in [1], it was extensively reported that CENTROG feature
technique gave highest accuracy results for vehicle type recognition at night time. Results from
these experiments showed an accuracy of 100% for the CENTROG technique in contrast to
92.7% for the CENTRIST technique as reported.
7. CONCLUSION
In conclusion, this paper proposed a feature-based technique for vehicle type recognition at both
day and night times. Initial categorisation was carried out to classify vehicles into day and night
time types using image histogram as features. Experiments conducted gave 100% recognition

54
accuracy. In order to recognise the vehicle types the proposed features were extracted by applying
Histogram Oriented Gradient on Census Transformed images and hence termed as CENTROG.
An SVM classifier was trained on the features obtained from the two datasets (day and night time
vehicles). The proposed technique was implemented and compared with the CENTRIST feature
technique. Experimental results showed that CENTROG outperformed the CENTRIST, recording
97.2% vs 95.6% (for day time) and 100% vs 92.7% (night time) recognition accuracies
respectively, thereby exhibiting a higher classification accuracy.
Future work would involve looking into identifying more categories, such as vans, tricycles and
motorcycles etc.
REFERENCES
[1] Martins E Irhebhude, Mohammad Athar Ali, and Eran A Edirisinghe. Pedestrian detection and
vehicle type recognition using centrog features for nighttime thermal images. In Intelligent Computer
Communication and Processing (ICCP), 2015 IEEE International Conference on, pages 407–412.
IEEE, 2015.
[2] Yoichiro Iwasaki, Masato Misumi, and Toshiyuki Nakamiya. Robust vehicle detection under various
environmental conditions using an infrared thermal camera and its application to road traffic flow
monitoring. Sensors, 13(6):7756– 7773, 2013.
[3] Martins E Irhebhude, Nawahda Amin, and Eran A Edirisinghe. View invariant vehicle type
recognition and counting system using multiple features. International Journal of Computer Vision
and Signal Processing, 6(1): 20-32, 2016.
[4] Khairi Abdulrahim and Rosalina Abdul Salam. Traffic surveillance: A review of vision based vehicle
detection, recognition and tracking. International Journal of Applied Engineering Research,
11(1):713–726, 2016.
[5] Noppakun Boonsim and Simant Prakoonwit. Car make and model recognition under limited lighting
conditions at night. Pattern Analysis and Applications, pages 1–13, 2016.
[6] Jakub Sochor, Adam Herout, and Jiri Havel. Boxcars: 3d boxes as cnn input for improved fine-
grained vehicle recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern
Recognition, pages 3006–3015, 2016.
[7] Jhonghyun An, Baehoon Choi, Kwee-Bo Sim, and Euntai Kim. Novel intersection type recognition
for autonomous vehicles using a multi-layer laser scanner. Sensors, 16(7):1123, 2016.
[8] Heikki Huttunen, Fatemeh Shokrollahi Yancheshmeh, and Ke Chen. Car type recognition with deep
neural networks. arXiv preprint arXiv:1602.07125, 2016.
[9] Ye Li, Bo Li, Bin Tian, and Qingming Yao. Vehicle detection based on the and-or graph for
congested traffic conditions. Intelligent Transportation Systems, IEEE Transactions on, 14(2):984–
993, 2013.
[10] Ehsan Adeli Mosabbeb, Maryam Sadeghi, and Mahmoud Fathy. A new approach for vehicle
detection in congested traffic scenes based on strong shadow segmentation. In Advances in Visual
Computing, pages 427–436. Springer, 2007.
[11] Ming Yin, Hao Zhang, Huadong Meng, and Xiqin Wang. An hmm-based algorithm for vehicle
detection in congested traffic situations. In Intelligent Transportation Systems Conference, 2007.
ITSC 2007. IEEE, pages 736–741. IEEE, 2007.
[12] S. Gupte, O. Masoud, R.F.K. Martin, and N.P. Papanikolopoulos. Detection and classification of
vehicles. Intelligent Transportation Systems, IEEE Transactions on, 3(1):37–47, 2002.
[13] PM Daigavane and PR Bajaj. Real time vehicle detection and counting method for unsupervised
traffic video on highways. International Journal of Computer Science and Network Security, 10(8),
2010.
[14] Chi-Chen Raxle Wang and J.-J.J. Lien. Automatic vehicle detection using local features;a statistical
approach. Intelligent Transportation Systems, IEEE Transactions on, 9(1):83–96, 2008.

55
[15] H.T.P. Ranga, M. Ravi Kiran, S. Raja Shekar, and S.K. Naveen Kumar. Vehicle detection and
classification based on morphological technique. In Signal and Image Processing (ICSIP), 2010
International Conference on, pages 45–48, 2010.
[16] Jun Yee Ng and Yong Haur Tay. Image-based vehicle classification system. arXiv preprint
arXiv:1204.2114, 2012.
[17] Celil Ozkurt and Fatih Camci. Automatic traffic density estimation and vehicle classification for
traffic surveillance systems using neural networks. Mathematical and Computational Applications,
14(3):187, 2009.
[18] Mehran Kafai and Bir Bhanu. Dynamic bayesian networks for vehicle classification in video.
Industrial Informatics, IEEE Transactions on, 8(1):100–109, 2012.
[19] Douglas Reynolds. Gaussian mixture models. Encyclopedia of Biometrics, pages 659–663, 2009.
[20] Zoran Zivkovic. Improved adaptive gaussian mixture model for background subtraction. In Pattern
Recognition, 2004. ICPR 2004. Proceedings of the 17th International Conference on, volume 2, pages
28–31. IEEE, 2004.
[21] Documentation OpenCV. Background subtraction, Accessed 23rd January 2014.
[22] Chris Stauffer and W Eric L Grimson. Adaptive background mixture models for real-time tracking. In
Computer Vision and Pattern Recognition, 1999. IEEE Computer Society Conference on., volume 2.
IEEE, 1999.
[23] Jianxin Wu, Christopher Geyer, and James M Rehg. Real-time human detection using contour cues.
In Robotics and Automation (ICRA), 2011 IEEE International Conference on, pages 860–867. IEEE,
2011.
[24] Ramin Zabih and John Woodfill. Non-parametric local transforms for computing visual
correspondence. In Computer VisionECCV’94, pages 151–158. Springer, 1994.
[25] Navneet Dalal and Bill Triggs. Histograms of oriented gradients for human detection. In Computer
Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society Conference on, volume
1, pages 886–893. IEEE, 2005.
[26] Irfan Riaz, Jingchun Piao, and Hyunchul Shin. Human detection by using centrist features for thermal
images. In International Conference Computer Graphics, Visualization, Computer Vision and Image
Processing. Citeseer, 2013.
[27] Chikahito Nakajima, Massimiliano Pontil, Bernd Heisele, and Tomaso Poggio. Full-body person
recognition system. Pattern recognition, 36(9):1997–2006, 2003.
[28] Sonka Milan, Hlavac Vaclav, and Boyle Roger. Image Processing Analysis, and Machine Vision.
Cengage Learning, Delhi, third edition, 2008.
[29] Christopher JC Burges. A tutorial on support vector machines for pattern recognition. Data mining
and knowledge discovery, 2(2):121–167, 1998.
[30] Marvin Smith, Joshua Gleason, Steve Wood, and Issa Beekun. Vehicle location by thermal images
features, Accessed 5th November 2014.
[31] Tom Fawcett. An introduction to roc analysis. Pattern recognition letters, 27(8):861–874, 2006.
[32] MD Thomas G. Tape. Interpreting diagnostic tests, retrieved 15th June, 2014.
AUTHORS
Martins Ekata Irhebhude obtained his tertiary and master degree education in Edo State,
Nigeria in 2003 and 2008 respectively. He is concluded his PhD research degree in 2015
with the Computer Science Department in Loughborough University, UK under the
supervision of Eran A. Edirisinghe PhD, a Professor of Digital Image Processing. Martins
is a staff of the Nigerian Defence Academy, Kaduna State, Nigeria since 2004 and currently
engages in teaching and research activities in and around the Defence Academy.
Research interests includes: object detection, people tracking, people re- identification, object
recognition and vision related researches.

56
Philip Oshiokhaimhele Odion got his first degree (BSc, Computer Science) in 1996 from
University of Benin, Benin-city, Edo State-Nigeria. He obtained his MSc, Computer Science
at Abubakar Tafawa Balewa University, Bauchi-Nigeria in 2006 and PhD, Computer Science
at Nigerian Defence Academy, Kaduna in 2014. Dr PO Odion is currently a Senior Lecturer
and Head of Department (HOD) of Computer Science at the Nigerian Defence Academy,
Kaduna. He has various local and international publications to his credit. His research interests are in
Software Engineering, Computer Networks and Artificial Intelligence.
Darius Tienhua Chinyio received the B. Ed. degree in Mathematics Education, and a Post
Graduate Diploma in Computer Science from Ahmadu Bello University, Zaria, Nigeria, in
1983 and 1986, respectively; and an M. Sc. in Computer Science from the University of
Lagos, Nigeria in 1991. He is currently working toward the Ph.D. degree in Computer
Science, Department of Computer Science, Nigerian Defence Academy, Kaduna, Nigeria;
under the supervision of Professor E. A. Onibere. His research interests include Computational Science,
Image Processing, and Networking.

AN INNOVATIVE RESEARCH FRAMEWORK ON INTELLIGENT TEXT DATA CLASSIFICATION SYSTEM USING GENETIC ALGORITHM

Recommended

More Related Content

What's hot (19)

Viewers also liked (19)

Similar to AN INNOVATIVE RESEARCH FRAMEWORK ON INTELLIGENT TEXT DATA CLASSIFICATION SYSTEM USING GENETIC ALGORITHM (20)

Recently uploaded (20)

AN INNOVATIVE RESEARCH FRAMEWORK ON INTELLIGENT TEXT DATA CLASSIFICATION SYSTEM USING GENETIC ALGORITHM