Alinteri J. of Agr. Sci. (2021) 36(2): 18-21 http://dergipark.gov.

tr/alinterizbd
e-ISSN: 2587-2249 http://www.alinteridergisi.com/
[email protected] DOI:10.47059/alinteri/V36I2/AJAS21111

RESEARCH ARTICLE

A Comprehensive Study on Intelligence System for Automatize Event

Tracker System Using Learning Method
Dr. Balakrishnan Natarajan1* • Dr.A. Vanitha2
1*Associate Professor, Department of Master of Computer Applications, Sona College of Technology, Salem, Tamil Nadu, India.
E-mail: [email protected]
2Assistant Professor, Department of Master of Computer Applications, Sona College of Technology, Salem, Tamil Nadu, India.
E-mail: [email protected]

ARTICLE INFO ABSTRACT

Article History: In image processing, the radical scheme is required to propose a model for extracting the
Received: 25.04.2021 required content from an image. It plays a critical position to offer significant facts and
Accepted: 02.06.2021 needs methods in various automation arenas. By keeping the way of a parting textual
Available Online: 12.07.2021 content from images has proposed via following the sparse matrix illustration, grouping text
components are based on heuristic rules and clustered into sentence generation. This paper
Keywords: directs a study on image analysis that inspects visual items as objects and different text
patterns. Logistic Regression, Linear Discriminant Analysis naïve Bayes Algorithm are used to
Code Book
predict the image forms. This proposed work promotes the learning algorithm called
Data Cleaning
Learning Vector Quantization Prediction Algorithm (LVQ Predict) is used to analysis the parts
Image Extraction
of the image. The features are extracted and classifies into printed and non-printed texts.
Learning Vector
Further, these texts are normalized and documented.
Non-printed Textures
Prediction Algorithm
Printed Textures

Please cite this paper as follows:

Dr. Natarajan, B. and Dr. Vanitha, A. (2021). A Comprehensive Study on Intelligence System for Automatize Event Tracker System Using
Learning Method. Alinteri Journal of Agriculture Sciences, 36(2): 18-21. doi: 10.47059/alinteri/V36I2/AJAS21111

Introduction
Pre-processing [6] is an essential steps to identify the
elements of an image that transforms e-image into a
collection of attributes a good way to be interpreted into
the OCR system. This technique consist the features of
grayscale methods, pixel into binary transformation,
thinning process to remove unwanted backgrounds, obtain
historical characteristics, segmentation and scalability
process. Those are extracting features and classifying into
further. In figure is stated as the following:

* Corresponding author: [email protected] Figure 1. Certificate Image Features Extraction and

Classification

18
 Dr. Natarajan, B. and Dr. Vanitha, A. (2021). Alınteri Journal of Agriculture Sciences 36(2): 18-21

A few researches [7- 8] that have been accomplished in approaches. There are three prominent components of the
binarization and segmenting might be reviewed as a propped model, MMA-SR is implemented into feature
recommendation of the technique used in the system. The extraction, multimodal attention and word prediction. This
threshold has set with respect to hue, bitmaps, and model has looked at conventional image captioning, OCR
segmentation range. based images are transformed into a spatial relationship of
OCR is the stage with the study of supervised learning the image that correlates the similarities, textures and
algorithm on machine learning that facilitate to understand patterns. Each entity is accessed into various objects and
the picture relies on characteristics, devise into classes with continues a historic repository associated with LSTM. This
highest accuracy from the image set [9-10]. In this model, muti-modularity simplifies and categorize the features are
figure 2 has described that has taken the samples of various defined. The final stage is a prediction of words by means of
images as input. It has six components that follow gray enhancing the probabilities of words are mapped with
scaling, binarization, segmenting, background removing, spatial relationship sets.
thinning and scaling which supports the feature extraction, Yuming He [2] focused generalized image knowledge
classifying into set of objects. The streamlined object can with the use of Deep-Learning based algorithm are
be normalized from the dictionaries whereas has historical efficiently worked with images on classification, detection
information that predicts word can be extracted and and segmentation. Its miles automatic to look into, analyze
provides meaningful information. the function are hidden in photographs with the aid of
repetitive stimulating guidelines some of the records-set.
Seelavathy, et al [3], It's far an elaborate mission due to
innovative movements of cellular digital camera beside by
manner of hand on shaking, transforming illumination at
hand over shade movement, and so forth. It is filtered out
from more icons are configured in this model which is
improvised the quality of transcription, increase the time of
responsiveness and more memory consumption has saved are
observed.
Nathiya N & Pradeepa K [4] has proposed a quick and
useful cropping algorithm is designed to extract multi
orientated textual content from an image. The enter picture
is first filtered with the related element method. Related
thing clustering is then used to identify candidate text areas
based totally on the most distinction. The frame of every
linked thing allows splitting the exceptional textual content
strings from every other. Then normalize candidate word
regions and decide whether every vicinity includes textual
content or now not. The size, skew, and shade of each
candidate may be envisioned from CCs, to expand a
text/non-textual content classifier for normalized snapshots.
on this strategies no longer only discover textual content, it
also extracts from the image and acknowledges the text in
phrases of storing the diagnosed phrases into a separate file
with the aid of incorporating numerous key upgrades over
traditional existing strategies to advise a unique CC
clustering-based totally scene textual content detection
approach, which subsequently ends in widespread overall
Figure 2. Image Classification Process performance improvement over the other competitive
methods.
DAR is a technique introduced in fully convolutional A unique textual content extraction approach [5] was
networks that focusing text and object localization to presented from GIF images. Graphical and document related
recognize the text by proposed language modelling images containing text and graphics additives are taken into
facilitated for handwritten images. In additional with this, consideration as 2D in which defines morphological traits.
signature are verified, document are categorized and The algorithm relies upon a sparse illustration framework
retrieved [11]. with as it should be selected discriminative over complete
dictionaries, each one offers sparse illustration over one sort
of signal and non-sparse representation over the opposite.
Related Work Separation of text and photographs additives is obtained
Jing Wang [1] applied a sentence decoder that gives a through selling sparse illustration of input pix in those two
technique to predict words through a multi-modularity dictionaries. Some heuristic guidelines are used for grouping
attention model that determine the features of the images. text additives into textual content strings in submit-
The comparison made between convention and OCR-based processing steps. The proposed approach overcomes the
19
 Dr. Natarajan, B. and Dr. Vanitha, A. (2021). Alınteri Journal of Agriculture Sciences 36(2): 18-21

hassle of touching among textual content and portraits.

Preliminary experiments display some promising effects on
special types of file.

Intelligence System for Automatize Event

Tracker System Using Learning Method
Enhancing the images is the challenging task is the real
scenario. The main objective is to improve the visibility of
the images and further, extract the various features of the Figure 3. LVPredict Architecture Diagram
images for predicting the required segment of the images.
The texts in these regions are extracted as printed and
There many techniques available to enhance the images
non-printed textures. These textures are analysed and
either by equalizing the pixel using the histogram, improving
duplicates are removed. Then, it is stored into the
the contrast, or applying the transformation to the features
documents as categorized.
of the images. Artificial Intelligence works in integrating the
The predictions on the images are made by defining the
human with the machine in human cognition, acquiring and
new instance (X) upon searching the codebook vectors for
calculating the events of processing. Many artificial
the K most instances. This first part of algorithm segregates
intelligence techniques are processing the symbolic
the image features. After classification are completed, the
reasoning in building the recognition and learning actions.
data mugging function is performed. To predict the
The machine learning solves the complex problems in a
duplication Euclidean distance is calculated. The similar
faster way of computing to yield best outcomes. Machine
images with new input are compared by this distance
Learning algorithms can able to recognize the speech to
measure. The Euclidean distance can be calculated by
text, sensing based outcome, effort estimations and lots
E(X,xi) = sqrt(sum(Xj – xij)^2)) (1)
more. Some of the machine learning algorithms that are From the equation (1), Euclidean distance E can be
used for predicting are linear regression, Logistic Regression, calculated by finding the square root of the summation of
Linear Discriminant Analysis Naïve Bayes and more. The the difference between the new point (Xj) and the existing
proposed system uses Learning Vector Quantization point xi.
algorithms (LVQ). LVQ is a supervised learning technique is The matched on the image part are removed and
used to predict the image parts. The proposed Learning
remaining part are extracted to store as documents.
Vector Prediction (LVPredict) algorithm initially, extracts
the features and classifies the images as program title,
program participants’ name, program dates and organizer’s Results and Discussion
details. Further, these details are normalized to reduce the The Learning Vector Quantization Prediction method
duplications in data store. (LVPredict) predicts by reading the codebook book data
The input image is divided into distinct regions and for randomly as input vectors. The vector instances are
each region reconstruction is defined. These regions are processed one at a time. The Learning algorithm with LV
classified and reproduced as a vector. The collection of Predict extracts the image features and avoids the
possible vectors are termed as code book of the quantifiers. duplication in an efficient manner. The image parts are
constructed as vectors in such a way to undergo
normalization process. After normalization process, the data
as the documents are stored in the data store efficiently.

Figure 4. Image Extraction Process

20
 Dr. Natarajan, B. and Dr. Vanitha, A. (2021). Alınteri Journal of Agriculture Sciences 36(2): 18-21

The Figure 4 is the explains the process of extracting

the images using the Learnng Algorithms. Here the text
content is sepeated as printed and non-printed textures.
Thes eclasifications are further analyzed, normalized and
stored as a document.

Conclusion
The Learning Vector Qubatization Prediction
(LVQPredict) is proposed to predict the images into textures.
These textures are clasified into printed and non-printed
text. Then the dupilcations are removed by finding the
Euclidean Distance Measure. Finally, the text are stores as
document for future accessing.

References
J. Wang, J. Tang and J. Luo, "Multimodal Attention with
Image Text Saptial Relationship for OCR-Based Image
Captioning," In ACM International Conference on
Multimedia, 2020.
Y. He, "Research on Text Detection and Recognition Based
on OCR Recognition Technology," In IEEE
international conference on Information Systems and
Computer Aided Education, 2020.
Sheelavathy K V and Priya Navneet, "Extracting Text from
the Picture by using OCR Technology," Intetrnational
Journal of Advanced Research in Computer Science,
vol. 11, no. Special, pp. 53-55, May 2020.
Nathiya N and Pradeepa K, "Optical Character Recognition
for scene text detection, mining and recognition," In
IEEE international Conference on Computational
Inteligence and Computing Research, 2014.
T. V. Hoang and S. Tabbone, "Text extraction from graphical
document images using saprse representation," In
IAPR Interanational Workshop on Document Analysis
System, June 2020.
N I Widiastuti and K E Dewi, "Document Image Extraction
System Design," IOP Conf. Series: Materials Science
and Engineering, Vol. 879, no. 1, 2020.
A. Fernández-Caballero, M. T. López and J. C. Castillo,
"Display text segmentation after learning best-fitted
OCR binarization parameters," Expert Systems with
Applications, vol. 39, no. 4, pp. 4032-4043, 2012.
A. Cheung, M. Bennamoun and N. W. Bergmann, "An Arabic
optical character recognition system using
recognition-based segmentation," Pattern
recognition, vol. 34, no. 2, pp. 215-233, 2001.
M. R. Phangtriastu, J. Harefa and D. F. Tanoto, "Comparison
between neural network and support vector machine
in optical character recognition," Procedia Computer
Science, vol. 116, pp. 351-357, 2017.
S. Naz, K. Hayat, M. I. Razzak, M. W. Anwar, S. A. Madani
and S. U. Khan, "The optical character recognition of
Urdu-like cursive scripts," Pattern Recognition, vol.
47, no. 3, pp. 1229-1248, 2014.
Cheng-Lin Liu, F. Gernot A, G. Venu and J. Lianwen, "Special
issue on deep learning for document analysis and
recognition," International Journal on Document
Analysis and Recognition (IJDAR), vol. 21, August
2018.
21