Detection of Anomalous Behaviour in an Examination Hall towards

Automated Proctoring

ABSTRACT
The paper proposes a workflow for the automatic detection of anomalous behavior in an
examination hall, towards the automated proctoring of tests in classes. Certain
assumptions about normal behavior in the context of proctoring exams are made.
Anomalies are behavior patterns that are relatively (and significantly) different. While not
every anomalous behavior may be cause for suspicion, the system is designed to
detection typical patterns for actions of concern such as discussions during an exam or
the turning around or the passing of notes, etc. This detection is based on features
computed using the textural features followed by a classifier search through annotated
patterns of pre- recorded clips to train the system for behavior that may cause concern.
While there may be false positives, the system is intended as a decision support system to
facilitate automatic proctoring of tests and deters malpractice. We have discussed about
various video analytics or video processing and image processing methods and tools
involved in surveillance model. We throughout the paper have walked through about the
various processes- preprocessing segmentation, classification, feature extraction and its
related video processing algorithm in sequential manner. The proposed model is
effective, efficient and requires relatively less processing power.
Introduction:

• The major problem that occurs in examination system is malpractices. This is

identified due to the absence of credible identity verification system for offline and
also for online examinations. In order to overcome the above problem researchers
have focused on the use of artificial techniques and use of biometrics. In the past
history work has been carried out on examination malpractices. Examination
malpractice is any form of an illegal act committed during the examination. There
are different forms of examination malpractice including copying from another
student’s test, getting notes to examination, plagiarism and impersonating another
student during a test. All these scenarios and many others give students an unfair
advantage. There may be many factors that cause examination malpractice like
physiological factors, societal value system, over emphasis on paper qualification
and poor learning facilities.

Existing Systems
• Fuzzy clustering
• Otsu method
• KNN classifier algorithm

Disadvantages

• Approximate result at the regulation of speed and direction

• It cannot work on the problems of scattering and at non co-ordinate
system

PROPOSED METHOD
• Haar-Features Algorithm
• AdaBoost Algorithm
• Principal Component Analysis(PCA)
Advantages
• Concept is easy to understand
• optimization Good for “noisy” environments
• It support multi objective

Block diagram:
Preprocessing
Image Pre-processing is a common name for operations with images at the lowest
level of abstraction. The preprocessing is the initial step for the process of video frames;
it includes conversion of image frames into grayscale image. The preprocessing of input
video is done to remove the noise and outliers, which are performed by Gaussian Filter.
Its input and output are intensity images. The aim of pre-processing is an improvement of
the image data that suppresses unwanted distortions or enhances some image features
important for further processing.Image restoration is the operation of taking a
corrupted/noisy image and estimating the clean original image. Corruption may come in
many forms such as motion blur, noise, and camera misfocus.Image Pre-processing is a
common name for operations with images at the lowest level of abstraction. Its input and
output are intensity images. The aim of pre-processing is an improvement of
the image data that suppresses unwanted distortions or enhances some image features
important for further processing.

Ada Boost
• Ada Boost, short for “Adaptive Boosting”, is the first practical boosting algorithm
proposed by Freund and Schapire in 1996. It focuses on classification problems
and aims to convert a set of weak classifiers into a strong one. The final equation
for classification can be represented as where f_m stands for the m_th weak
classifier and theta_m is the corresponding weight. 

Ada-boost or Adaptive Boosting combines multiple classifiers to increase the accuracy of

classifiers. AdaBoost is an iterative ensemble method. AdaBoost classifier builds a strong
classifier by combining multiple poorly performing classifiers so that you will get high
accuracy strong classifier. The basic concept behind Adaboost is to set the weights of
classifiers and training the data sample in each iteration such that it ensures the accurate
predictions of unusual observations. Any machine learning algorithm can be used as base
classifier if it accepts weights on the training set. Adaboost should meet two conditions:
The classifier should be trained interactively on various weighed training examples. In
each iteration, it tries to provide an excellent fit for these examples by minimizing
training error. Boosting is a general ensemble method that creates a strong classifier from
a number of weak classifiers. This is done by building a model from the training data,
then creating a second model that attempts to correct the errors from the first model.
Models are added until the training set is predicted perfectly or a maximum number of
models are added. Adaboost was the first really successful boosting algorithm developed
for binary classification. It is the best starting point for understanding boosting.

BACKGROUND SUBSTRACTION

Background subtraction, also known as foreground detection, is a technique in the

fields of image processing and computer vision wherein an image's foreground is
extracted for further processing (object recognition etc.). Background subtraction is a
widely used approach for detecting moving objects in videos from static cameras. The
rationale in the approach is that of detecting the moving objects from the difference
between the current frame and a reference frame, often called "background image", or
"background model". Background subtraction is mostly done if the image in question is a
part of a video stream. Background subtraction provides important cues for numerous
applications in computer vision, for example surveillance tracking or human poses
estimation.
Hardware Requirements
• system
• 4 GB of RAM
• 500 GB of Hard disk

SOFTWARE REQUIREMENTS:

⮚ Python
⮚ Open-CV

REFERENCES

[1] D. Weinland, R. Ronfard, and E. Boyer, “Free viewpoint action recognition using
motion history”, volumes. Computer Vision and Image Understanding, 104(2), 2006

[2] M. Blank, L. Gorelick, E. Shechtman, M. Irani, and R. Basri, “Actions as space-

timeshapes”. In Proc. ICCV, 2005

[3] Dehghani, Alireza, and Alistair Sutherland. "A novel interest-point-based

background subtraction algorithm." ELCVIA Electronic Letters on Computer Vision
and Image Analysis 13.1 (2014).

[4] Ben-Musa, Ahmad Salihu, Sanjay Kumar Singh, and Prateek Agrawal.
"Suspicious Human Activity Recognition for Video Surveillance System." (2014).

[5] Yong, Lu, and He Dongjian. "Video-based detection of abnormal behavior in the
examination room." Information Technology and Applications (IFITA), 2010
International Forum on. Vol. 3. IEEE, 2010.