International Journal of Computational Intelligence Research

ISSN 0973-1873 Volume 13, Number 5 (2017), pp. 745-765

© Research India Publications
http://www.ripublication.com

A Review of Detection and Tracking of Object from

Image and Video Sequences

Mukesh Tiwari
Research Scholar, Department of Electronics & Communication, University Institute
of Technology, Rajiv Gandhi Technical University, Bhopal – 462033 M.P. (India)

Dr. Rakesh Singhai

Professor, Department of Electronics & Communication, University Institute of
Technology, Rajiv Gandhi Technical University, Bhopal – 462033 M.P. (India).

Abstract

Object detection and tracking is one of the critical areas of research due to
routine change in motion of object and variation in scene size, occlusions,
appearance variations, and ego-motion and illumination changes. Specifically,
feature selection is the vital role in object tracking. It is related to many real
time applications like vehicle perception, video surveillance and so on. In
order to overcome the issue of detection, tracking related to object movement
and appearance. Most of the algorithm focuses on the tracking algorithm to
smoothen the video sequence. On the other hand, few methods use the prior
available information about object shape, color, texture and so on. Tracking
algorithm which combines above stated parameters of objects is discussed and
analyzed in this research. The goal of this paper is to analyze and review the
previous approach towards object tracking and detection using video
sequences through different phases. Also, identify the gap and suggest a new
approach to improve the tracking of object over video frame.

Keywords: Object tracking, object recognition, statistical analysis, object

detection, background subtraction, performance analysis, optical flow

1. INTRODUCTION
Recently, there is an advance of miniaturization and lower the cost of cameras have
preferred the implementation of large-scale networks of the camera. This increasing
number of cameras could permit novel signal processing applications which employ
multiple sensors in extensive areas. Object tracking is the novel procedure for
discovering moving objects beyond time by utilizing the camera in video sequences
746 Mukesh Tiwari and Dr. Rakesh Singhai

(Kothiya and Mistree, 2015). Their main aim is to relate the target objects as well as
the shape or features, location of the objects in successive video sequences.
Subsequently, the object classification and detection are essential for object tracking
in computer vision application. Additionally, the tracking is the first step towards
locating or detects the moving object in the frame. Followed by this, detected object
could be divided as swaying tree, birds, human, and vehicles and so on. Though, in
image processing approach object tracking using video sequences, is a challenging
task. Furthermore, several issues appear ascribed to occlusion of the object to scene,
object to object, complex object motion, real-time processing requirements as well as
the improper shape of the object.
However, this tracking has a large number of benefits, few of them are traffic
monitoring, robot vision, surveillance and security and video communication, public
areas like underground stations, airports, mass events and animation (Kim, 2007;
Lowe, 2004; Ojha and Sakhare, 2015; Yilmaz et al., 2006). Thus, the particular
application needs optimal trade-off among computing, communication, and accuracy
over the network. The revenue related to computing and communication relies on the
amount and type of cooperation executed among cameras for data collection,
dispensing and processing to confirm decisions and to reduce the estimation errors
and ambivalence.
Subsequently, this tracking can be explained as the procedure of determining the
orientation of object across the time as the object moves throughout a scene. This is
posting importance in the arena of computer vision because of expansion of high-
powered computers and the growing need for automated surveillance systems, and it
is broadly applied for applications namely automated surveillance, robotics
monitoring, human-machine interface, motion-based recognition, vehicle navigation,
traffic monitoring and video indexing. A substantial number of such applications
require reliable tracking methods which meet real-time restrictions and are
challenging and complex with respect to changes of object movement, scale and
appearance, illumination of scene and occlusion. The results of tracking could be
impacted by the disparity of one among the parameters. Due to tackle the above-
explained issues and others in object tracking numerous approaches have been
proposed (Yilmaz et al., 2006). In this object tracking application, target object could
be determined as anything which is engaging for analysis. In addition, moving objects
tracking is one of the major tasks in computer vision and broadly applied in industrial
vision, intelligent transport systems and visual surveillance (Comaniciu et al., 2003,
2000). In the recent years, Video surveillance has widely adopted to monitor the
security sensitive areas include highways, borders, department stores, banks and
crowded public places. The development in computing power, the infrastructure of
high-speed network and accessibility of large-capacity storage devices cover the way
for inexpensive, multi-sensor video surveillance systems. Keeping a track on the
moving object is a critical task.
The capability of machines to identify the suspicious object and further identify their
activities in a specific environment is an important part of permitting a machine to
interact with humans in effective and easy manner. The current approach for
analyzing and detecting the suspicious object usually needs exceptional markers
A Review of Detection and Tracking of Object from Image and Video Sequences 747

connected to the suspicious object that prevents the extensive technology

application. In this paper, to study as wll as analyze the previous approach towards
object tracking using video sequences through different phases.
Three key steps in video analysis are discussed as follows:
1. Identification of targeted object in moving sequence.
2. Object tracking based on one frame to another frame.
3. Tracking of the object from camera to camera.

2. LITERATURE REVIEW
In the previous study most of them have concentrated towards Object detection (Ben
Ayed et al., 2015; Najva and Bijoy, 2016; Ramya and Rajeswari, 2016; Risha and
Kumar, 2016; Shen et al., 2013; Soundrapandiyan and Mouli, 2015; Viswanath et al.,
2015) ,Object tracking (Bagherpour et al., 2012; Coşkun and Ünal, 2016; Foytik et al.,
2011; Lee et al., 2012; Poschmann et al., 2014; Weng et al., 2013; Yilmaz et al., 2006;
Zhang et al., 2016) and Object recognition (Chakravarthy et al., 2015; Elhariri et al.,
2015; Gang et al., 2010; Ha and Ko, 2015; Nair et al., 2011) for tracking the object
using video sequences. These are discussed as follows. The basic flow diagram of an
object tracking shown in figure 1.

Video Object Object Object

sequence Detection Recognition Tracking

Figure 1 The Basic flow diagram of Object tracking

Studies related to object detection

The detection of an object in video sequence plays a significant role in many
applications. Specifically as video surveillance applications (Amandeep and Goyal,
2015). The different types of object detection are shown in figure 2.
Object
Detection

Frame Optical Flow Background

Difference Subtraction

Non-
Recursive Recursive

Figure 2 Types of object detection method

748 Mukesh Tiwari and Dr. Rakesh Singhai

The previous studies related to object detection is discussed as follows:

Shen et al. (2013) proposed a novel hierarchical moving target detection method
based on spatiotemporal saliency. Further, they achieved the refined detection results
using temporal and spatial saliency information. The experimental results show that
this approach identifies moving substances in the airborne video with high accuracy
and efficiency. Additionally, this method does not have the effect of time delay when
compared with HMI technique. However, this method evaluated object locations in all
the video frames as self-sufficiently, false alarms which are inevitable.
Guo et al. (2012) suggested object detection approach for tracking the objects in video
frames. The simulation result shows this technique was effective and accurateness,
robust for generic object classes’ detection with good performance. Further needs to
focus towards increase classification accurateness in real-time object recognition.
Ben Ayed et al. (2015) proposed a method for detection of text data based on a texture
in video frames by big data analytics. The video frames are decomposed into various
fixed size blocks and these blocks are analyzed using har wavelet transform
technique. Further, they used a neural network to classify the text and non-text blocks.
However, this study needs to concentrate on extracting the regions towards remove
the noisy regions as well as exclude the text like sections.
Viswanath et al. (2015) suggested and modeled the approach using non-panoramic
background modeling. By the use of this approach, they modeled the entire picture
element with one Spatio-temporal Gaussian. The simulations result shows this method
able to identify the moving substances with fewer false alarms. However, this method
fails once the adequate features are not obtainable from the section.
Soundrapandiyan and Mouli (2015) suggested a novel and adaptive method for
pedestrian detection. Further, they separated the foreground objects from the
background by image pixel intensities. Subsequently, they used high boost filter for
enhancing the foreground edges. The efficacy of the proposed method is evident from
the subject evaluation results as well as objective evaluation with around 90% of
pedestrian’s detection rate compared to the other single image existing methods. In
future, they planned to improve the performance of the method with higher detection
rate and low false positives on par with sequence image methods.
Ramya and Rajeswari (2016) suggested a modified frame difference method which
uses the correlation between blocks of current image and background image to
categorize the pixels as foreground and background. The blocks in the current image
which are highly correlated with the background image are considered as background.
For the other block, the pixel-wise comparison is made to categorize it as foreground
or background. The experiments conducted proved this approach improves the frame
difference method particularly as finding accuracy with speed. However, this study
needs to focus towards other information available in the blocks such as shape and
edge can be used to improve the detection accuracy.
Risha and Kumar (2016) suggested an optic flow with the morphological operation
for object detection in video. Further applied morphological operation towards
A Review of Detection and Tracking of Object from Image and Video Sequences 749

obtaining clear moving target image. This study only concentrated on static camera.
So need to focus on moving the camera as well as identify multiple objects in video
frames.
Najva and Bijoy (2016) proposed a model for detection and classification of objects in
videos by combining Tensor features with SIFT approach towards classifying the
detected objects using Deep Neural Network(DNN). The DNN capable of handling
large higher dimensional data with billions of parameters as like human brain.
Simulation results obtained illustrate that the proposed classifier model produces more
accurate results than the existing methods, which combines both SIFT and tensor
features for feature extraction and DNN for classification.

Table 1 Comparative Study of Object Detection technique

Object Detection Basic Principle Computa Accur Comments
Method tional acy
Time
Temporal Pixel-wise Low High Easy to implement (Chate et al.,
Differencing Subtraction of 2012; Mohan and Resmi, 2014)
Current & Sensitive to dynamic changes
Background frame (Haritaoglu et al., 2000)
Needs background frame with
still objects (Mohan and Resmi,
2014)
Backgro Frame Current frame is Low to Moder Simplest background Subtraction
und Differen subtracted from Moderate ate to (Aldhaheri and Edirisinghe, 2014;
Subtract cing background frame High Haritaoglu et al., 2000)
ion Cannot be used for real-time
applications (Mohan and Resmi,
2014)
Approxi Simple subtraction Low to Moder No need for adequate
mate between median moderate ate background modeling (Aldhaheri
Median frame & test frame and Edirisinghe, 2014)
Requires a buffer with recent
pixel values (Aldhaheri and
Edirisinghe, 2014)
Running Based on Gaussian Moderate Moder Much suitable for real-time
Gaussian probability density to high ate applications (Aldhaheri and
Average function of pixels Edirisinghe, 2014)
Statistical calculations consumes
more time
Mixture Based on Moderate Moder Low memory requirement
of multimodal to high ate to (Zhiqiang et al., 2006)
Gaussian distribution high Cannot cope up with objects as
well as noise(Tao Zhang et al.,
2010)
Optical Uses optical flow Moderate High This approach offers entire
Flow distribution to high moving data(Krishna et al., 2011)
characteristics of however require more
pixels of object calculations
750 Mukesh Tiwari and Dr. Rakesh Singhai

Studies related to object tracking

The object tracking is the term which used to identify the moving object position as
well as tracking them from video sequences (Balasubramanian et al., 2014). The
tracking method is classified into three types such as kernel, point and silhouette
based tracking (Yilmaz et al., 2006). Compared to silhouette method, existing most of
them have focused on kernel-based method due to high accuracy with less
computational cost. However, the point tracking method has less computational cost
with reduce in accuracy (Weng et al., 2013). The various types of object tracking
techniques are shown in figure 3.

Object
Tracking

Kernel Silhoutte
Point
Tracking Tracking
Tracking

Simple
Template Contour
Matching Matching
Kalmal
Filter

Mean Shift
Method
Shape
Matching
Particle
Filter

Support
Multiple Vector
Hypothese Machine
s

Layering
Based
Tracking

Figure 3 Types of object tracking approach

The previous studies related to object tracking method are discussed as follows:
Sarkar et al. (2012) developed color information method for identification of skin in
image occurred from FERET especially for mouth and eye region detection. This
technique requires less computational cost as well as applied to video sequences.
However, this method incapable towards identifying the small face images which
away from the camera.
A Review of Detection and Tracking of Object from Image and Video Sequences 751

Weng et al. (2013) proposed an algorithm for improving the performance of natural
feature selection in the real world. Further, they used to speed up robust features
(SURF) for features extraction from live mobile camera image and recognition. These
extracted features are calculated using pose matrix through Homography approach.
The simulation result shows this algorithm tracked and recognized the object from
natural features in easy, speedy with suitable way. However, its speed and accuracy
need to be improved.
Zhang et al. (2016) proposed an approach by combining frame difference and non-
parametric method for video analysis traceability. The simulation result proved this
approach performance was better than the traditional frame difference and GMM.
Further, it can able to remove the noise from a background which gives us the ability
to detect the moving object more precisely in the applications such as food and
agriculture related product traceability analysis. However, this study needs to enhance
the capability of traceability system and supply the visual supply chain for the
common user to ensure the safety.
Coşkun and Ünal (2016) suggested camshift technique towards track the object from
the video sequences. Further, they demonstrated this approach successful carried out
in mobile platform even with the change in object size and illumination. The
drawback observed in this proposed technique fails to perform for input video with
full occlusion. Houssineau et al. (2016) suggested a parameterization based disparity
space for non-rectified camera networks, extended to moving objects, and integrated
into a Bayesian multi-object tracking with sensor calibration technique. The
performance of the obtained framework has been demonstrated for camera calibration
on simulated with real data, underlying the problems of single-object localization and
tracking, as well as for multi-object tracking. Further, they planned to enhance the
proposed method towards other multi-object filters and a comparative study of these
approaches for camera-based tracking as well as for camera calibration.
Oiwa et al. (2016) suggested probabilistic background model towards tracking the
object from video sequences. The simulation results show the accuracy and
effectiveness of this method high compared to previous technique. However, this
study needs to concentrate on higher speed as well as improve the accuracy of object
tracking.
Mohammed and Morris (2014) suggested a color-based technique which was the
combination of accruing and normalizing histograms towards object tracking under
different conditions based on a mobile device. This technique was easy to use and
robust against varying illuminations. However, this technique fails to identify the
entire region of symbols due to the extreme camera view.
Aggarwal et al. (2006) suggested a novel technique which was the combination of
motion estimation and background subtraction for object tracking using video
sequences. The system mainly focused on four scenarios such as interpolation,
identify the object, subtract the background and object selection. However, this study
needs to focus more towards full occlusions of video sequences, multiple object
tracking, fast camera motion and unsupervised object tracking.
752 Mukesh Tiwari and Dr. Rakesh Singhai

Fatima et al. (2013) suggested image segmentation approach for track objects of
interests through specifying the color intensities. A minimum distance classifier
approach is used for object classification. They attain the object tracking through
specifying the object centroids in all the video frames. The simulation results proved
this technique was more efficient for contextual approximation. However, this study
needs to focus towards implementing segmentation which would work with occluded
images with multiple objects and computationally more efficient.
Lecumberry and Pardo (2005) proposed an algorithm for semi-automatic object
tracking in videos by various features with probabilistic relaxation method. The
proposed algorithm performs effectively for object tracking specifically for object
borders is smooth and accurate. When combined with sophisticated methods it
performs effectively for object tracking. The accuracy of the borders of the tracked
objects depends on the power of discrimination of selected features as well as the
appearance of new objects and/or background. However, this study needs to focus
towards improving the accuracy of this algorithm, particularly as rigid objects.

Table 2 Comparative Study of Object Tracking Methods

Object tracking Algorithm used Computatio Accur Comments
method nal time acy
Point Kalman Kalman filtering Low toModer This approach applicable to track point
Tracking Filter algorithm Moderate ate even in noisy images (Javed and Shah,
2002)
Distributed State variables (Gaussian)
(Ali and Zafar, 2009; Javed and Shah,
2002)
Particle Recursive Bayes Moderate to High Good results for occultation and
Filter filtering High complex background (Hu et al., 2004)
Not advisable for real-time applications
due to big calculations (Athanesious and
Suresh, 2012; Hu et al., 2004)
Multiple MHT algorithm Low Low to Adapt new object as well as exists
Hypothesi modera existing object(Blackman, 2004)
s tracking te High computation in memory and
time(Blackman, 2004)
Kernel Simple Matching region Low to Low Capable of dealing with partial
Tracking template of interest in moderate occlusion (Cucchiara et al., 2003)
matching video Require equivalent model for each
region of interest for each image
(Cucchiara et al., 2003)
Mean shift Expression & Low Moder Can be used for real-time applications
method location of ate due to less calculations (Tao Zhang et
object; optimal al., 2010)
gradient decline Iterations get into local maximum
easily (Tao Zhang et al., 2010)
Support Positive & Moderate Moder Can handle single image and partial
vector negative training ate occlusions (Avidan, 2004)
A Review of Detection and Tracking of Object from Image and Video Sequences 753

machine values Need physical initialization and training

(Avidan, 2004)
Layering Shape Moderate Moder Track multiple objects and full
based representation ate to occlusion (Wei and Qin, 2007)
tracking using intensity high Require parametric models of each
pixel (Wei and Qin, 2007)
Silhouett Contour Gradient Descent Moderate Moder Object Shape is Implicitly modeled
e matching Algorithm ate to (Karasulu, 2010)
tracking high Requires time for state space estimation
(Karasulu, 2010)
Shape Hough High High Less sensitive towards variance of
matching Transform appearance
Need to enhance the performance
(Karasulu, 2010)

Table 3 Comparative Study of Object Classification Methods

Object Computa Accuracy Comments
Classification tional
Method Time
Shape Based Low Moderate to Pattern matching algorithm was applied. However, it's not
high worked at dynamic situation as well as unable towards
control internal movements (Javed and Shah, 2002)
Motion Based High Moderate struggles towards recognizing on-moving humanoid
(Javed and Shah, 2002)
Color based High High Delivers expenditure of added computation time with
improved quality(Hu et al., 2004)
Texture Based High High They described the color distribution using Gaussian
Mixture Model in image sequence as well as segment
background and objects of the image(Hu et al., 2004)

Studies related to object recognition

Gang et al. (2010) developed a kernel locality preserving projections (KLPP) towards
improving the accuracy and separable description of the objects. The simulation
results proved, this approach more appropriate for space object recognition mainly
considering changes of viewpoints. However, this study needs to concentrate on
improving the accuracy level towards recognizing with limited trained objects as
reference. Also, test the effectiveness of this approach on images of other objects
models.
Nair et al. (2011) developed a technique for a combination of face recognition,
tracking and detection approach towards identifying the individual faces. The
simulation result shows this method increases the level of accuracy in recognizing and
tracking faces and applicable to the real-time application. However, it fails in varying
illumination conditions as well as the existence of lesser background in face
recognition which limits the system performance.
754 Mukesh Tiwari and Dr. Rakesh Singhai

Zhang and Jiang (2014) suggested regression based kernel technique on behalf of
identifying multiview objects as well as approximating their poses. The simulation
results proved obtained results are improved recognition, while comparative analyses
with state-of-the-arts. Further, they authenticated the robustness and efficiency of this
approach. However, the existing research needs to focus more attention on resolving
issues of object space recognition
Chakravarthy et al. (2015) suggested and estimated a technique for video sequence
stream processing. Further, they demonstrated the various types of situations which
were based on arable and relational illustrations. However, this study fails to express
complex situations. So, needs to focus towards accepting feature vectors as well as
bounding boxes towards identifying the temporal and spatial computations.
Ha and Ko (2015) proposed vision-based shadow gesture recognition method for
interactive projection systems. This method only splits the shadow area through
merging the binary image with an input image using a learning algorithm that isolates
the background from the input image. The developed approach isolates or
differentiates shadow of a hand in based on convex hull, moment and defect in each
region. After that, isolated the hand shadows by the convex hull, defect, and moment
of each region. However, the robustness of this processing is not always assured. It
means that at some conditions the gesture recognition fails, and unexpected results
would happen.
Elhariri et al. (2015) discussed Random Forests (RF), Linear Discriminant Analysis
(LDA) classifiers, Support Vector Machines (SVMs).Furthermore, for edge detection
segmentation approach has been utilized in this research. Subsequently extracted the
features using morphological operations. The simulation result shows the LDA and
SVM provide better performance. However, the input dataset as elderly needs to be
involved.

2.1 Statistical methods of tracking

Feris (2000) suggested a statistical skin-color model towards detecting the various
regions of candidate’s face as well as applied template matching techniques to define
presence or absence of a face in particular region. They tracked the face of the
candidate by lip corners, the location of pupils and nostrils location in the tracked
image sequence. However, this study needs to improve the system robustness and
accuracy of detection as well as tracking the facial features.
Foytik et al. (2011) proposed a Kalman filter for low-level recognition of
differentiating between various objects. Then they analyzed the face recognition at
various subspace formed from the big database and analyzed through Adaptive
MPCA (Modular Principal Component Analysis). The simulation results proved this
approach has more accuracy and able to make recognized faces with average time.
Bagherpour et al. (2012) developed a novel framework for tracking of upper parts’ of
the body using KLT (Kanade-Lucas-Tomasi) and Kalman filter for motion estimation.
In the developed model limbs and motion of the body is estimated using Kalman filter
A Review of Detection and Tracking of Object from Image and Video Sequences 755

for accurate prediction of limb position. The experimental result shows that the
developed model is compatible with the indoor environment for variation in object
pose, background clutter and for various scenes. This study can further concentrate on
improvisation of tracking results.
Lee et al. (2012) investigated different occlusion scenario and performed tracking
under six different video simulation methods. They evaluated the performance using
SFDA (Sequence Frame Detection Accuracy). Furthermore, they demonstrated mean
shift, particle and Kalman filtering for evaluating tracking performance. Additionally,
they found that for arbitrary movement of the object Particle Filter (PF) fails to
perform effectively.
Kim (2007) objects are randomly chosen by a user are tracked using SIFT features
and a Kalman filter. Specifically, they concentrated on tracking human, car, or pre-
learned objects. The objects are accumulated, exploited the learning to successfully
track objects even when the objects missing for some frames. However, this study
needs to focus on higher resolution with finding the location of stationary objects.
Nagendran et al. (2014) proposed a method for effectively tracking moving objects in
videos. They used affine transformation for stabilizing the video. Then extract these
features using frame selection. Further, they used Kalman filter and Gaussian mixture
model for tracking the moving objects. However, this study needs to concentrate on
reduction of computational time as well as increasing recognition for various
categories.
Poschmann et al. (2014) developed a PF approach using fusion technique for
increasing adaptive tracking robustness. This research comparatively analyzed the
various variants and demonstrated the feasibility of applying a framework for a real-
world scenario. The major difficulty identified in this research is the threshold for
learning is crucial which will be either too high or too low. Another issue identified is
based on video, threshold is not updated whether bad or none. The stated problem can
be overcome by exploitation of adaptive threshold feasibility in proposed approach or
else need to find an alternate way to this challenge.
Mei et al. (2015) proposed a LAD (least absolute deviation) learning method based on
a multitasking and multiview technique for tracking. The proposed approach uses PF
for effective object tracking. The proposed approach is implemented under four
different features of objects like color histogram, intensity, LBP (Local binary
patterns) and HOG (Histogram of Oriented Gradients). Further, this research is
examined under several challenging situations like noise availability in real-world,
synthetic noisy sequence, availability of sequence in public and complete tracking of
available data sets. The simulated results demonstrate that proposed method was
provided the advantage of multiview data handling and task outlier. Further, the
proposed approach exhibits superior performance for comparative examination of
existing tracking methods.
Chandrajit et al. (2016) suggested a feature based method towards track the multiple
moving objects in surveillance video sequence is proposed. This proposed method is
assessed quantitatively using the precision and recall accuracy metrics. Further,
756 Mukesh Tiwari and Dr. Rakesh Singhai

comparative evaluation with related works has been carried out to exhibit the efficacy
of the proposed method. However, this study needs to focus on occlusion handling
strategy and shadow elimination for increase the tracking results.
Moreno-Garcia et al. (2010) developed a motion-based algorithm for accurate
prediction of moving objects shapes in the video sequence. By using fuzzy logic
sequence segmentation and tracking performance were enhanced which improves
uncertainty of work based on permanency value caused by impairment in computer
vision. However, this study needs to focus on improving the Analysis phase in order
to make it much more descriptive. In order to conquer the previous image for local
search, tracking phase will be modified in the fuzzy region or else need to create a
method to fit with the fuzzy region.

3. MATERIALS AND METHODS

The present study reviewed 74 research papers sharing the same context of the present
study. The papers are selected from the search of comprehensive literature performed
using the IEEE transactions, reputed international journals such as Elsevier, Springer,
ACM, and papers through Google search. With the review and reference section from
previous research papers, individual papers were selected for the study of previous
literature. The selected studies support the video-based object detection, tracking and
recognition with performance analysis. After careful analysis of previous papers, 48
articles that serve to be potentially useful for the present study were selected for the
research.

4. RESULTS AND DISCUSSION

Shen et al. (2013) proposed a novel hierarchical moving target detection method.
However, the drawback of the proposed method is the unavoidable cases of false
alarms since frame by frame detection of object location is estimated.
Soundrapandiyan and Mouli (2015) suggested a method for improving the
performance with higher detection rate and low false positives on par with sequence
image methods. Ramya and Rajeswari (2016) suggested a modified frame difference
method which uses the correlation between blocks of current image and background
image to categorize the pixels as foreground and background. However, this study
needs to focus towards other information available in the blocks such as shape and
edge can be used to improve the detection accuracy. Sarkar et al. (2012) developed a
technique for face detection using color information availability. The model is
potential enough to detect faces from images that are of low-resolution or from
surveillance video frames. Furthermore, the proposed model also identifies face
alignment in the images. Hence, image registration is facilitated, and the registered
images are used for future uses.
Weng et al. (2013) enhanced the performance of tracking for real-world objects using
naturally formulated approach. However, its speed and accuracy need to be improved
through replacing SURF algorithm with the different computationally efficient
algorithm. Poschmann et al. (2014) incorporated particle fusion approach with other
A Review of Detection and Tracking of Object from Image and Video Sequences 757

existing techniques for enhancing tracking performance which facilitate the adaptive
tracking of the objects. The threshold for learning faces is a crucial task wherein if the
threshold is too high or low, then there arise chances of bad or no updates which
depends on the characteristics of the video. Hence, a necessity persists to examine the
possibility of possessing an adaptive threshold or identify new methods to resolve this
issue. Zhang et al. (2016) developed a novel algorithm for improving video analysis
traceability by combining frame difference and non-parametric techniques. However,
this study needs to enhance the capability of traceability system and supply the visual
supply chain for the common user to ensure the safety of food and agriculture related
products. A conventional camshaft approach is used by Coşkun and Ünal (2016) to
track objects. However, with full occlusion, the proposed model will fail. Gang et al.
(2010) devised 3d Objects based identification based on kernel locality preserving
projections (KLPP).In a similar manner, Nair et al. (2011) developed a system for
detection and recognition of various pose of the face also it tracks face efficiently.
The proposed model also extends the face detection algorithm to track and recognized
individual profile faces. However, the study limits its scope to face detection
algorithm whereas there is a dire need to focus on skin segmentation algorithm which
is used to create a face mask. The face mask creation mechanism is an exceptional
algorithm wherein with the creation of face based on the pose; background could be
neglected from the face training system. Additionally, pose estimation module is
required for the determination of pose and suitable Eigenspace should be selected
rather than the use of cascades. Elhariri et al. (2015) proposed an object recognition
system which is based on the use of SVMs, LDA classifiers, RF and Histogram of
Oriented Gradients. However, the previous study lacks focus on certain features
which when applied with other machine learning classifiers achieves better accuracy
of recognition. Objects that are faced previously are included as the present datasets
does not include them. Additionally, faces could be utilized for the extension of the
proposed system’s applicability.
Many researchers have concentrated namely on Particle filter, Mean shift, and
Kalman filter. Several works have been proposed to combine the prospects of these
trackers in order to achieve better results of tracking. For instance, previous studies of
Li et al. (2010) and Zhao et al. (2009) have combined Mean-shift tracking and
Kalman filtering whereas Tang and Zhang (2011) devised a combined model of PF
with Mean Shift tracking. One of the most explored objects tracking issues in
computer vision is the handling of occlusion. Existing literature works have argued
that occlusion handling is achieved through Particle filter, Kalman filter and Mean
Shift tracking methods. Though the fusion of these methods is accurate, they are
tested only on particular videos (Lee et al., 2012). It is also deemed that face detection
methods imparting neural network-based approaches are exceptionally accurate and
are computationally intensive in nature. Hence, there is a lack of possibilities in
implementing neural approaches in real-time applications (Sarkar et al., 2012).
Visual tracking is facilitated by the least squares (LSs) criterion which is the use of
Euclidean distance to calculate approximately tracking of sparse representation (Xue
Mei and Haibin Ling, 2011; Zhang et al., 2013). It is deemed that if Gaussian data
distribution is involved, then the performance of LS criterion is efficient. The reason
758 Mukesh Tiwari and Dr. Rakesh Singhai

is attributed to the properties of smoothness and differentiability and solutions are

mostly based on gradient methods (Nesterov, 2013). However, real tracking instances
are prone to noise which includes cases of Laplace noise, background clutter, and salt-
and-pepper noise. In such cases, the methods of LS- criterion are into serious
degradation and hence accurate estimation of noise could not be performed. Through
the analysis of existing researchers, Guan et al. (2012) and Harter (1974) claim that
least absolute deviation (LAD) has better efficiency in estimation of noise than LS,
especially in cases where heavy-tailed noise prevails. On the contrary, tracking issues
imply different data representations which are categorized based on intensity (Mei et
al., 2007), color (Collins et al., 2005), edge (Kwon and Le, 2010), wavelet (Jepson et
al., 2003), and texture. Tracking performance could be improved significantly if
different sources of information are considered since with these characteristics,
tracking could be facilitated (Badrinarayanan et al., 2007; Du and Piater, 2008; Kwon
and Le, 2010; Weifeng Liu and Dacheng Tao, 2013). Though multiple views are to be
considered, another issue is the ways of integrating data representations and a model
to examine independence and mutual dependence. Research by Zhang et al. (2013)
reported that for visual tracking applications features are not possible to implement as
individual task hence it needs to be an exhibit in all sequential tasks. Furthermore,
outlier tasks also exist very often. Outlier tasks do not share the features that are
common with the majority of tasks. Similarly, Zhang et al. (2013) and Xue Mei and
Haibin Ling (2011) utilized the feature of intensity to model target’s appearance
change. The intensity appearance model with L1 minimization shares robustness
towards partial OCC and other tracking issues (Xue Mei and Haibin Ling, 2011).
Even though this developed model is sensitive to track non-rigid objects since shape
deformation is more evident with these objects.
The present study has reviewed 74 articles, from which 19 from IEEE digital library,
7 from an international conference, 8 were from springer digital library, 6 from
Elsevier, 6 from Science direct, and 28 from Google scholar (Figure 4). Of these
articles, the following aspects are addressed: Object detection, tracking, recognition
and statistical methods of tracking specifically as a video sequence.

IEEE

International Conference

Springer

Elsevier

Science Direct

Google Scholar

Figure 4 Number of research article referred

A Review of Detection and Tracking of Object from Image and Video Sequences 759

5. CONCLUSION
In this paper, review on different object detection, tracking, recognition techniques,
feature descriptors and segmentation method which is based on the video frame and
various tracking technologies. This approach used towards increase the object
detection with new ideas. Furthermore, tracking the object from the video frames with
theoretical explanation is provided in bibliography content. The bibliography content
is the most significant contribution of research since it will lead to a new area of
research. We have identified and discussed the limitation/future scope of various
methods. Also, we have noted some methods which give accuracy but have high
computational complexity. Specifically, the statistical methods, background
subtraction, temporal differencing with the optical flow was discussed. However,
these technique needs to concentrate towards handling sudden illumination changes,
darker shadows and object occlusions (Susar and Dongare, 2015).

6. FUTURE SCOPE
 Design and simulation of complex video sequence and test them using same
tracking algorithm. In the potential scenario, occlusion is used for an object with
the same color for the moving objects or else using bigger occlusion with longer
occlusion time. Increasing the number of the object help to identify the efficiency
and functionality of the tracking algorithm.
 Weight parameters are needed to be added for individual intensity levels of each
pixel. In an image, if an intensity value is assigned as foreground based on the
current frame then it has less probability that foreground also has similar pixel
coordinate so that BG weightage for the pixel is set to the minimum than the
initial value. Through adding weightage lower than the initial value provides the
advantage of removing the old pixel value with least probability rather than the
evolved scene.
 Need to focus towards enhancing the variance data of each channel based on the
Mahalanobis distance calculation. By this, can able to adopt a change in the rapid
scene through Euclidean distance algorithm.

REFERENCES
[1] Aggarwal, A., Biswas, S., Singh, S., Sural, S. & Majumdar, A.K., 2006.
Object Tracking Using Background Subtraction and Motion Estimation in
MPEG Videos, in 7th Asian Conference on Computer Vision. Springer-
Verlag Berlin Heidelberg, pp. 121–130. doi:10.1007/11612704_13
[2] Aldhaheri, A.R. & Edirisinghe, E.A., 2014. Detection and Classification of a
Moving Object in a Video Stream, in: Proc. of the Intl. Conf. on Advances in
Computing and Information Technology. Institute of Research Engineers and
Doctors, Saudi Arabia, pp. 105–111. doi:10.3850/ 978-981-07-8859-9_23
760 Mukesh Tiwari and Dr. Rakesh Singhai

[3] Ali, S.S. & Zafar, M.F., 2009. A robust adaptive method for detection and
tracking of moving objects, in: 2009 International Conference on Emerging
Technologies. IEEE, pp. 262–266. doi:10.1109/ICET.2009.5353164
[4] Amandeep, Goyal, M., 2015. Review: Moving Object Detection Techniques.
Int. J. Comput. Sci. Mob. Comput. 4, 345 – 349.
[5] Athanesious, J. & Suresh, P., 2012. Systematic Survey on Object Tracking
Methods in Video. Int. J. Adv. Res. Comput. Eng. Technol. 1, 242–247.
[6] Avidan, S., 2004. Support vector is tracking. IEEE Trans. Pattern Anal.
Mach. Intell. 26, 1064–1072. doi:10.1109/TPAMI.2004.53
[7] Badrinarayanan, V., Perez, P., Le Clerc, F. & Oisel, L., 2007. Probabilistic
Color and Adaptive Multi-Feature Tracking with Dynamically Switched
Priority Between Cues, in: 2007 IEEE 11th International Conference on
Computer Vision. IEEE, pp. 1–8. doi:10.1109/ICCV.2007.4408955
[8] Bagherpour, P., Cheraghi, S.A. & Bin Mohd Mokji, M., 2012. Upper body
tracking using KLT and Kalman filter. Procedia Comput. Sci. 13, 185–191.
doi:10.1016/j.procs.2012.09.127
[9] Balasubramanian, A., Kamate, S., & Yilmazer, N., 2014. Utilization of robust
video processing techniques to aid efficient object detection and tracking.
Procedia Comput. Sci. 36, 579–586. doi:10.1016/j.procs.2014.09.057
[10] Ben Ayed, A., Ben Halima, M., & Alimi, A.M., 2015. MapReduce-based text
detection in big data natural scene videos. Procedia Comput. Sci. 53, 216–
223. doi:10.1016/j.procs.2015.07.297
[11] Blackman, S.S., 2004. Multiple hypotheses tracking for multiple target
tracking. IEEE Aerosp. Electron. Syst. Mag. 19, 5–18.
doi:10.1109/MAES.2004.1263228
[12] Chakravarthy, S., Aved, A., Shirvani, S., Annappa, M., & Blasch, E., 2015.
Adapting Stream Processing Framework for Video Analysis. Procedia
Comput. Sci. 51, 2648–2657. doi:10.1016/j.procs.2015.05.372
[13] Chandrajit, M., Girisha, R., & Vasudev, T., 2016. Multiple Objects Tracking
in Surveillance Video Using Color and Hu Moments. Signal Image Process.
An Int. J. 7, 15–27. doi:10.5121/sipij.2016.7302
[14] Chate, M., Amudha, S., & Gohokar, V., 2012. Object Detection and tracking
in Video Sequences. ACEEE Int. J. Signal Image Process. 3.
[15] Collins, R.T., Yanxi Liu, & Leordeanu, M., 2005. Online selection of
discriminative tracking features. IEEE Trans. Pattern Anal. Mach. Intell. 27,
1631–1643. doi:10.1109/TPAMI.2005.205
[16] Comaniciu, D., Ramesh, V. & Meer, P., 2003. Kernel-based object tracking.
IEEE Trans. Pattern Anal. Mach. Intell. 25, 564–577.
doi:10.1109/TPAMI.2003.1195991
A Review of Detection and Tracking of Object from Image and Video Sequences 761

[17] Comaniciu, D., Ramesh, V. & Meer, P., 2000. Real-time tracking of non-
rigid objects using mean shift, in: IEEE International Proceedings of the
Computer Vision and Pattern Recognition. IEEE Computer Society, DC, pp.
142–149.
[18] Coşkun, M. & Ünal, S., 2016. Implementation of Tracking of a Moving
Object Based on Camshift Approach with a UAV. Procedia Technol. 22,
556–561. doi:10.1016/j.protcy.2016.01.116
[19] Cucchiara, R., Grana, C., Piccardi, M. & Prati, A., 2003. Detecting moving
objects, ghosts, and shadows in video streams. IEEE Trans. Pattern Anal.
Mach. Intell. 25, 1337–1342. doi:10.1109/TPAMI.2003.1233909
[20] Du, W. & Piater, J., 2008. A Probabilistic Approach to Integrating Multiple
Cues in Visual Tracking, in: Computer Vision – ECCV 2008. Springer Berlin
Heidelberg, Berlin, Heidelberg, pp. 225–238. doi:10.1007/978-3-540-88688-
4_17
[21] Elhariri, E., El-Bendary, N., Hassanien, A.E. & Snasel, V., 2015. An
Assistive Object Recognition System for Enhancing Seniors Quality of Life.
Procedia Comput. Sci. 65, 691–700. doi:10.1016/j.procs.2015.09.013
[22] Fatima, H., Shah, S.I.A., Jamil, M., Mustafa, F. & Nadir, I., 2013. Object
Recognition, Tracking and Trajectory Generation in Real-Time Video
Sequence. Int. J. Inf. Electron. Eng. 3, 639–642.
doi:10.7763/IJIEE.2013.V3.395
[23] Feris, R.S., 2000. Detection and Tracking of Facial Features in Video
Sequences, in: Lecture Notes in Artificial Intelligence. Springer-Verlag
Berlin Heidelberg, pp. 197–206.
[24] Foytik, J., Sankaran, P. & Asari, V., 2011. Tracking and recognizing multiple
faces using Kalman filter and ModularPCA. Procedia Comput. Sci. 6, 256–
261. doi:10.1016/j.procs.2011.08.047
[25] Gang, M., Zhiguo, J., Zhengyi, L., Haopeng, Z. & Danpei, Z., 2010. Full-
viewpoint 3D space object recognition based on kernel locality preserving
projections. Chinese J. Aeronaut. 23, 563–572. doi:10.1016/S1000-
9361(09)60255-7
[26] Guan, N., Tao, D., Luo, Z. & Shawe-Taylor, J., 2012. MahNMF: Manhattan
Non-negative Matrix Factorization.
[27] Guo, L., Liao, Y., Luo, D. & Liao, H., 2012. Generic Object Detection Using
Improved Gentleboost Classifier. Phys. Procedia 25, 1528–1535.
doi:10.1016/j.phpro.2012.03.272
[28] Ha, H. & Ko, K., 2015. A method for image-based shadow interaction with
virtual objects. J. Comput. Des. Eng. 2, 26–37.
doi:10.1016/j.jcde.2014.11.003
762 Mukesh Tiwari and Dr. Rakesh Singhai

[29] Haritaoglu, I., Harwood, D. & Davis, L.S., 2000. W/sup 4/: real-time
surveillance of people and their activities. IEEE Trans. Pattern Anal. Mach.
Intell. 22, 809–830. doi:10.1109/34.868683
[30] Harter, W.L., 1974. The Method of Least Squares and Some Alternatives:
Part I. Int. Stat. Rev. / Rev. Int. Stat. 42, 147. doi:10.2307/1403077
[31] Houssineau, J., Clark, D.E., Ivekovic, S., Lee, C.S. & Franco, J., 2016. A
Unified Approach for Multi-Object Triangulation, Tracking and Camera
Calibration. IEEE Trans. Signal Process. 64, 2934–2948.
doi:10.1109/TSP.2016.2523454
[32] Hu, W., Tan, T., Wang, L. & Maybank, S., 2004. A Survey on Visual
Surveillance of Object Motion and Behaviors. IEEE Trans. Syst. Man
Cybern. Part C (Applications Rev. 34, 334–352.
doi:10.1109/TSMCC.2004.829274
[33] Javed, O. & Shah, M., 2002. Tracking and Object Classification for
Automated Surveillance, in: Proceeding ECCV ’02 Proceedings of the 7th
European Conference on Computer Vision-Part IV. Springer-Verlag London,
UK, pp. 343–357.
[34] Jepson, A.D., Fleet, D.J. & El-Maraghi, T.F., 2003. Robust online appearance
models for visual tracking. IEEE Trans. Pattern Anal. Mach. Intell. 25, 1296–
1311. doi:10.1109/TPAMI.2003.1233903
[35] Karasulu, B., 2010. Review and Evaluation of Well-Known Methods for
Moving Object Detection and Tracking in Videos. J. Aeronaut. Sp. Technol.
4, 11–22.
[36] Kim, Y.M., 2007. Object Tracking in a Video Sequence.
[37] Kothiya, S. V. & Mistree, K.B., 2015. A review on real-time object tracking
in video sequences, in: 2015 International Conference on Electrical,
Electronics, Signals, Communication and Optimization (EESCO). IEEE, pp.
1–4. doi:10.1109/EESCO.2015.7253705
[38] Krishna, M.T.G., Ravishankar, M. & Babu, D.R.R., 2011. Automatic
detection and tracking of moving objects in complex environments for video
surveillance applications, in: 2011 3rd International Conference on
Electronics Computer Technology. IEEE, pp. 234–238.
doi:10.1109/ICECTECH.2011.5941596
[39] Kwon, J. & Le, K.M., 2010. Visual tracking decomposition, in: Proc. IEEE
Conf. Comput. Vis. Pattern Recognit. Portland, OR, USA, pp. 1269–1276.
[40] Lecumberry, F. & Pardo, A., 2005. Semi-Automatic Object Tracking in
Video Sequences, in: XI Congreso Argentino de Ciencias de La
Computación.
A Review of Detection and Tracking of Object from Image and Video Sequences 763

[41] Lee, B.Y., Liew, L.H., Cheah, W.S. & Wang, Y.C., 2012. Measuring the
effects of occlusion on kernel based object tracking using simulated videos.
Procedia Eng. 41, 764–770. doi:10.1016/j.proeng.2012.07.241
[42] Li, Y.H., Pang, Y.G., Li, Z.X. & Liu, Y.L., 2010. An intelligent tracking
technology based on Kalman and Mean shift algorithm. ICCMS 2010 - 2010
Int. Conf. Comput. Model. Simul. 1, 107–109. doi:10.1109/ICCMS.2010.76
[43] Lowe, D.G., 2004. Distinctive Image Features from Scale-Invariant
Keypoints. Int. J. Comput. Vis. 60, 91–110.
doi:10.1023/B:VISI.0000029664.99615.94
[44] Mei, X., Hong, Z., Prokhorov, D. & Tao, D., 2015. Robust Multitask
Multiview Tracking in Videos. IEEE Trans. Neural Networks Learn. Syst.
26, 2874–2890. doi:10.1109/TNNLS.2015.2399233
[45] Mei, X., Zhou, S.K. & Porikli, F., 2007. Probabilistic Visual Tracking via
Robust Template Matching and Incremental Subspace Update, in:
Multimedia and Expo, 2007 IEEE International Conference on. IEEE, pp.
1818–1821. doi:10.1109/ICME.2007.4285026
[46] Mohammed, A.D. & Morris, T., 2014. A Robust Visual Object Tracking
Approach on a Mobile Device, in: IFIP International Federation for
Information Processing. pp. 190–198.
[47] Mohan, A.S. & Resmi, R., 2014. Video image processing for moving object
detection and segmentation using background subtraction, in: 2014 First
International Conference on Computational Systems and Communications
(ICCSC). IEEE, pp. 288–292. doi:10.1109/COMPSC.2014.7032664
[48] Moreno-Garcia, J., Rodriguez-Benitez, L., Fernández-Caballero, A. & López,
M.T., 2010. Video sequence motion tracking by fuzzification techniques.
Appl. Soft Comput. 10, 318–331. doi:10.1016/j.asoc.2009.08.002
[49] Nagendran, A., Dheivasenathipathy, N., V. Nair, R. & Sharma, V., 2014.
Recognition and Tracking Moving Objects Using Moving Camera in
Complex Scenes. Int. J. Comput. Sci. Eng. Appl. 4, 31–36.
doi:10.5121/ijcsea.2014.4203
[50] Nair, B.M., Foytik, J., Tompkins, R., Diskin, Y., Aspiras, T. & Asari, V.,
2011. Multi-pose faces recognition and tracking system. Procedia Comput.
Sci. 6, 381–386. doi:10.1016/j.procs.2011.08.070
[51] Najva, N. & Bijoy, K.E., 2016. SIFT and Tensor-Based Object Detection and
Classification in Videos Using Deep Neural Networks. Procedia Comput. Sci.
93, 351–358. doi:10.1016/j.procs.2016.07.220
[52] Nesterov, Y., 2013. Gradient methods for minimizing composite functions.
Math. Program. 140, 125–161. doi:10.1007/s10107-012-0629-5
764 Mukesh Tiwari and Dr. Rakesh Singhai

[53] Oiwa, D., Fukui, S., Iwahori, Y., Nakamura, T. & Bhuyan, M.K., 2016.
Tracking with Probabilistic Background Model by Density Forests, in: Proc.
of IEEE/ACIS. IEEE, Okayama, Japan, pp. 391–396.
[54] Ojha, S. & Sakhare, S., 2015. Image processing techniques for object
tracking in video surveillance- A survey, in: 2015 International Conference
on Pervasive Computing (ICPC). IEEE, pp. 1–6.
doi:10.1109/PERVASIVE.2015.7087180
[55] Poschmann, P., Huber, P., Rätsch, M., Kittler, J. & Böhme, H.-J., 2014.
Fusion of Tracking Techniques to Enhance Adaptive Real-time Tracking of
Arbitrary Objects. Procedia Comput. Sci. 39, 162–165.
doi:http://dx.doi.org/10.1016/j.procs.2014.11.025
[56] Ramya, P. & Rajeswari, R., 2016. A Modified Frame Difference Method
Using Correlation Coefficient for Background Subtraction. Procedia Comput.
Sci. 93, 478–485. doi:10.1016/j.procs.2016.07.236
[57] Risha, K.P. & Kumar, A.C., 2016. Novel Method of Detecting Moving
Object in Video. Procedia Technol. 24, 1055–1060.
doi:10.1016/j.protcy.2016.05.235
[58] Sarkar, R., Bakshi, S. & Sa, P.K., 2012. A Real-time Model for Multiple
Human Faces Tracking from Low-resolution Surveillance Videos. Procedia
Technol. 6, 1004–1010. doi:http://dx.doi.org/10.1016/j.protcy.2012.10.122
[59] Shen, H., Li, S., Zhu, C., Chang, H. & Zhang, J., 2013. Moving object
detection in aerial video based on spatiotemporal saliency. Chinese J.
Aeronaut. 26, 1211–1217. doi:10.1016/j.cja.2013.07.038
[60] Soundrapandiyan, R. & Mouli, P.V.S.S.R.C., 2015. Adaptive Pedestrian
Detection in Infrared Images Using Background Subtraction and Local
Thresholding. Procedia Comput. Sci. 58, 706–713.
doi:10.1016/j.procs.2015.08.091
[61] Susar, R. & Dongare, M., 2015. Moving Object Detection, a Succinct
Review. Int. J. Adv. Res. Comput. Commun. Eng. 4, 334–336.
doi:10.17148/IJARCCE.2015.41277
[62] Tang, D. & Zhang, Y.J., 2011. Combining mean-shift and particle filter for
object tracking, in: Proceedings - 6th International Conference on Image and
Graphics, ICIG 2011. IEEE, pp. 771–776. doi:10.1109/ICIG.2011.118
[63] Tao Zhang, Zaiwen Liu, Xiaofeng Lian & Xiaoyi Wang, 2010. Study on
moving-objects detection technique in video surveillance system, in: 2010
Chinese Control and Decision Conference. IEEE, pp. 2375–2380.
doi:10.1109/CCDC.2010.5498797
A Review of Detection and Tracking of Object from Image and Video Sequences 765

[64] Viswanath, A., Kumari, R. & Senthamilarasu, V., 2015. Background

Modelling from a Moving Camera. Procedia - Procedia Comput. Sci. 58,
289–296. doi:10.1016/j.procs.2015.08.023
[65] Wei, L. & Qin, W.J.L., 2007. Study on Moving Object Tracking Algorithm in
Video Images, in: IEEE Conference on Electronic Measurement and
Instruments, ICEMI 8th International Conference. IEEE, pp. 810–816.
[66] Weifeng Liu & Dacheng Tao, 2013. Multiview Hessian Regularization for
Image Annotation. IEEE Trans. Image Process. 22, 2676–2687.
doi:10.1109/TIP.2013.2255302
[67] Weng, E.N.G., Khan, R.U., Adruce, S.A.Z., & Bee, O.Y., 2013. Objects
Tracking from Natural Features in Mobile Augmented Reality. Procedia -
Soc. Behav. Sci. 97, 753–760. doi:10.1016/j.sbspro.2013.10.297
[68] Xue Mei & Haibin Ling, 2011. Robust Visual Tracking and Vehicle
Classification via Sparse Representation. IEEE Trans. Pattern Anal. Mach.
Intell. 33, 2259–2272. doi:10.1109/TPAMI.2011.66
[69] Yilmaz, A., Javed, O., & Shah, M., 2006. Object tracking: A survey. ACM
Comput. Surv. 38, 13–es. doi:10.1145/1177352.1177355
[70] Zhang, H. & Jiang, Z., 2014. Multi-view space object recognition and pose
estimation based on kernel regression. Chinese J. Aeronaut. 27, 1233–1241.
doi:10.1016/j.cja.2014.03.021
[71] Zhang, J., Cao, J. & Mao, B., 2016. Moving Object Detection Based on Non-
parametric Methods and Frame Difference for Traceability Video Analysis.
Procedia Comput. Sci. 91, 995–1000. doi:10.1016/j.procs.2016.07.132
[72] Zhang, T., Ghanem, B. & Liu, S., Ahuja, N., 2013. Robust Visual Tracking
via Structured Multi-Task Sparse Learning. Int. J. Comput. Vis. 101, 367–
383. doi:10.1007/s11263-012-0582-z
[73] Zhao, J., Qiao, W. & Men, G.Z., 2009. An approach based on mean shift and
Kalman filter for target tracking under occlusion. Proc. 2009 Int. Conf. Mach.
Learn. Cybern. 4, 2058–2062. doi:10.1109/ICMLC.2009.5212129
[74] Zhiqiang, W., Xiaopeng, J. & Peng, W., 2006. Real-time is moving object
detection for video monitoring systems. J. Syst. Eng. Electron. 17: 731–736.
doi:10.1016/S1004-4132(07)60007-3.
766 Mukesh Tiwari and Dr. Rakesh Singhai