Professional Documents
Culture Documents
Suspicious Object Detection With Alarm Notification For Security Personnel
Suspicious Object Detection With Alarm Notification For Security Personnel
Suspicious Object Detection With Alarm Notification For Security Personnel
Abstract
This study aimed to develop a “suspicious object detection with alarm notification for security
personnel”. The main objective of this study was to track or detect suspicious objects, analyze
whether it was abandoned, and send a notification to the security personnel. A prototype methodology
was used in the development of the system which employed an image processing methodology to
analyze the video captured by the cameras. The image processing applied the background extraction,
foreground detection, thresholding algorithm, noise reduction algorithm, tracking and detecting
algorithm. The system was evaluated by 10 security personnel and 10 I.T. experts using the ISO
25010 criteria in terms of functional suitability, reliability, performance, efficiency, operability,
security, compatibility, maintainability, and portability. The overall result of the evaluation was
excellent which implied that the system met the standard criteria in system development.
for automatic detection and analysis of suspicious store information. The development of reel-to-reel
objects, especially in places that prohibited no parking media enabled the recording of surveillance footage.
for vehicles and those suspicious bystanders. Image These systems required magnetic tapes to be changed
capturing capability had contributed to the recognition manually which was a time-consuming, expensive and
of image knowledge. With a system that could track unreliable process, with the operator having to
suspicious objects, it could prevent future accidents or manually thread the tape from the tape reel through the
problems; a system that could coordinate with security recorder onto an empty take-up reel. Closed-circuit
personnel for early notification and detection. television was used as a form of pay-per-view theatre
television for sports such as professional boxing and
Research Objectives professional wrestling from 1964 through 1970, the
Indianapolis 500 automobile race.
Generally, the study aimed to develop suspicious
object detection with alarm notification for security The closed-circuit television concept was very
personnel. Specifically, the study aimed to: (1) manage essential in coming up with this study. It gave an idea
live video feed, recorded videos, and archived videos; about video cameras transmitting those captured
(2) detect and track suspicious objects; (3) send images across another user or medium. The concept of
notifications to security personnel once a suspicious that was integrated in this system where it had a digital
object is detected. camera and could be broadcasted or accessed by
authorized personnel through different networks. It
was very useful concept to improve the current CCTV
Literature Review
system that we had now.
Security is freedom from, or resilience against, Computer vision is a field of artificial intelligence that
potential harm (or other unwanted coercive change) trained computers to interpret and understand the
caused by others. Beneficiaries (technical referents) of visual world. Using digital images from cameras and
security may be of persons and social groups, objects videos and deep learning models, machines could
and institutions, ecosystems, or any other entity or accurately identify and classify objects — and then
phenomenon vulnerable to unwanted change. Public react to what they “saw.”
security was the function of governments that ensured
the protection of citizens, persons in their territory, Early experiments in computer vision took place in the
organizations, and institutions against threats to their 1950s, using some of the first neural networks to
well-being – and to the prosperity of their communities detect the edges of an object and to sort simple objects
[19]. into categories like circles and squares. In the 1970s,
the first commercial use of computer vision interpreted
The concept of suspicious object detection with alarm typed or handwritten text using optical character
notification for security personnel was based on an recognition. This advancement was used to interpret
idea of safety and security that people are currently written text for the blind [21]. As the internet matured
experiencing. The problem of undetected or suspicious in the 1990s, making large sets of images available
objects is one of the major problems that people are online for analysis, facial recognition programs
facing. With the proposed system, it could have an flourished. These growing data sets helped make it
opportunity to safeguard establishment owners with possible for machines to identify specific people in
security features. photos and videos.
Closed-circuit television (CCTV) The effects of these advances on the computer vision
field had been astounding. Accuracy rates for object
Video surveillance is the use of video cameras to identification and classification had gone from 50
transmit a signal to a specific place on a limited set of percent to 99 percent in less than a decade — and
monitors. It differed from broadcast television in that today’s systems were more accurate than humans at
the signal was not openly transmitted, though it may quickly detecting and reacting to visual inputs. The
employ point-to-point (P2P), point-to-multipoint study was anchored on this concept of computer
(P2MP), or mesh wired or wireless links [20]. The vision, this system had computers and digital cameras.
earliest video surveillance systems involved constant The system captured images or footages from the
monitoring because there was no way to record and cameras for analysis called image processing
procedure and then it detected or tracked objects. improvements in computing power and storage.
The way image segmentation identifies the set of Automatic thresholding is a great way to extract useful
pixels is very similar to the approach of this study, information encoded into pixels while minimizing
wherein it identified the separated segments of images background noise. This is accomplished by utilizing a
for tracking and detection purposes. feedback loop to optimize the threshold value before
converting the original grayscale image to binary. The
Image Classification idea is to separate the image into two parts: the
background and foreground. The thresholding for
Given a set of images that are all labelled with a single image processing helped extract data useful for
category, it is imperative to predict these categories for analyzing.
a novel set of test images and measure the accuracy of
the predictions. There are a variety of challenges The idea and the algorithm process had a major role in
associated with this task including viewpoint variation, creation of this study because it was implemented
scale variation, intra-class variation, image purely to achieve the objectives of the study. By the
deformation, image occlusion, illumination conditions help of thresholding segmentation process, the system
and background clutter [25]. was able to extract the entire background to be
compared to the foreground of the image. It helped in
The most popular architecture which has been used for analyzing the different objects that had been added to
image classification is Convolutional Neural Networks the foreground for detection and tracking purposes.
(CNNs). A typical use case for CNNs was where you
feed the network images and the network classified the Otsu's method
data. CNNs tended to start with an input “scanner”
which was not intended to parse all the training data at In computer vision and image processing, Otsu's
once. For example, to input an image of 100 x 100 method, named after Nobuyuki Otsu (Ōtsu Nobuyuki)
pixels, you wouldn’t want a layer with 10,000 nodes. is used to perform automatic image thresholding. In
the simplest form, the algorithm returns a single
This input data is then fed through convolutional intensity threshold that separates pixels into two
layers instead of normal layers. Each node only classes, foreground and background [27]. This
concerns itself with close neighboring cells. These threshold is determined by minimizing intra-class
convolutional layers also tends to shrink as they intensity variance, or equivalently, by maximizing
became deeper, mostly by easily divisible factors of inter-class variance. Otsu's method- a one-dimensional
the input. Besides these convolutional layers, they also discrete analog of Fisher's Discriminant Analysis is
often featured pooling layers. Pooling is a way to filter related to Jenk’s optimization method and is
out details. A commonly found pooling technique is equivalent to a globally optimal k-means.
max pooling, where, 2x2 pixels are taken and passed
on the pixel with the most amount of a certain The algorithm exhaustively searches for the threshold
attribute. that minimizes the intra-class variance defined as a
weighted sum of variances of the two classes.
The image classification process which has a similar
implementation towards this study, both of these Clustering Methods
process analyze images by identifying certain
differences in a given data. With that, the classification The K-means algorithm is an iterative technique used
and identification to exploit the necessary data for to partition an image into K clusters. The basic
further analysis were also applied in this study. algorithm are as follows: 1. Pick K cluster centers,
either randomly or based on some heuristic method,
Thresholding (image processing) for example K-means++. 2. Assign each pixel in the
image to the cluster that minimized the distance
Thresholding is the simplest method of segmenting between the pixel and the cluster center. 3. Re-
image. From a grayscale image, thresholding could be compute the cluster centers by averaging all of the
used to create binary images. The simplest pixels in the cluster. 4. Repeat steps 2 and 3 until
thresholding methods replace each pixel in an image convergence was attained (i.e. no pixels change
with a black pixel. If the image intensity or a white clusters). K-means clustering was a method of vector
pixel is greater than that constant. In the example quantization, originally from signal processing,
image on the right, these results in the dark tree popular for cluster analysis in data mining. K-means
becoming completely black, and the white snow clustering aimed to partition in observations into k
clusters in which each observation belonged to the vision including image retrieval and video
cluster with the nearest mean, serving as a prototype of surveillance.
the cluster. This resulted in a partitioning of the data
space into Voronoi cells. The concept of object detection is to analyze images or
objects present in the image. After the analysis, it
This idea of Otsu's method or clustering method has a compares that kind of object to the data set in the
similarity to this study through the concept of system. That system could then mark or track the
identifying the foreground and the background by particular object that has been added to the scene or
applying segmentation and thresholding to one image. the object. The idea of this approach is very much
However, the current study applied this kind of similar to this study, it could mark an object that has
approach by analyzing a separate image for the been added and detect an object similar to object
foreground and the background image. detection. However, the difference between the two is
that the object detection is more likely to be a machine
Edge Detection learning approach compared to this study. It is more on
image processing approach and tracking. But the idea
Edge detection includes a variety of mathematical of machine learning using object detection approach
methods that aimed at identifying points in a digital could also be applied to this study. That is why it
image at which the image brightness changed sharply would be considered as an enhancement or
or, more formally have discontinuities. The points at recommendation for this study.
which image brightness change sharply are typically
organized into a set of curved line segments termed Survey on Abandoned Object Detection using
edges [28]. Image Processing
The same problem of finding discontinuities in one-
Survey on Abandoned Object detection using Image
dimensional signals was known as step detection and
Processing by Dhanashri Adalinge (2007). The
the problem of finding signal discontinuities over time
scenario for the proposed system used Foreground
was known as change detection. Edge detection was a
detection for image processing as a method in
fundamental tool in image processing, machine vision
identifying abandoned object: then, it used blob
and computer vision, particularly in the areas of
algorithm that track moving entity. With that proposed
feature detection and feature extraction.
or design of the system, it could detect an object by
In finding the object and identifying the object shape separating the foreground and focusing on the moving
accurately, it is needed to have a definite data objects [30].
representation of that object. The edge detection
Foreground Detection
approach is similar to this study. It needs to identify
the object from the background image and by that it
Foreground detection is a technique of image
needed to determine the precise shape of the object for
processing used to separate objects and people from
separation and analysis similar to this study.
the overall background. In case of crowded places, the
Object detection background is the infrastructure of the area, while the
foreground would be the people or independent
Object detection is a computer technology related to structures. With this approach, it is important to clear
computer vision and image processing that dealt with the data from the objects. It only detected objects on
detecting instances of semantic objects of a certain the foreground layer once it separates and identifies
class (such as humans, buildings, or cars) in digital the background layer information [34]; then after that
images and videos [29]. process has taken place, it would start tracking the
object.
It traditionally has applications in video and real-world
interactions where observations are made following Blob Detection
initial object detection. Now, it is crucial to
autonomous driving systems such as self-driving In computer vision, blob detection methods aim at
vehicles from companies like Uber and Tesla. detecting regions in a digital image that differed in
properties such as brightness or color compared to
Well-researched domains of object detection includes surrounding regions. Informally, a blob is a region of
face detection and pedestrian detection. Object an image in which some properties are constant or
detection has applications in many areas of computer approximately constant. All the points in a blob could
be considered in some sense to be similar to each camera for detection is similar to this study. The study
other. The most common method for blob detection is aimed to detect and track objects. However, in this
convolution [31]. The survey on abandoned object study it only tracked an object if there was motion or
detection using image processing was similar to the movement in the footage. While the proposed study
study in terms of tracking and detection of object. The was able to detect and track objects added to the scene
system however, needed to apply thresholding to and it was not merely focused on the movement alone.
extract the background while this study did not
mention it. Both of the system applied foreground Active Video-Based Surveillance System
analysis in order to track object and send notification
for security purposes. Those kinds of approaches were Foresti (2005) [33] proposed a method that included
very useful to come up with this study. change method detection for fixed cameras. A variety
of change detection methods were developed for fixed
A Survey on Visual Surveillance of Object Motion cameras. The system also has background updating.
and Behaviors To minimize errors in the background change
detection process due to noise effects, illumination,
A Survey on Visual Surveillance of Object Motion and and/or environmental changes, advanced visual-based
Behaviors by Weiming Hu (2004). The scenario for surveillance systems applied background updating
the proposed system was for object detection and procedures. The system had changed detection
behaviors, it used environment modeling - an active methods for mobile cameras. Since the camera
construction and updating of the environmental model movement induces an apparent motion in all of the
as a method. Then it applied motion segmentation that image pixels, the application of standard CD
aimed to detect regions corresponding to moving techniques resulted in poor motion detection and
objects such as vehicles and humans. It included therefore could not be used.
background subtraction, a popular method for motion
segmentation, especially under those situations with a The concept and approaches of this study was similar
relatively static background and temporal differencing to this study. It used a fixed or stable camera so that
that used pixel-wise differences between two or three the system would compare the new objects added to
consecutive frames in an image sequence to extract the scene then it would start tracking the object. It has
moving regions. Temporal differencing was very the capability as well to update the background to
adaptive to dynamic environments, but generally does avoid errors in detecting objects. In this study, the
a poor job of extracting all the relevant pixels, e.g., page had to refresh manually to have a new
there may be holes left inside moving entities. background reference; however, it was not capable of
recording such scenes and it did not have a security
Nearly every visual surveillance system starts with features yet.
motion detection. Motion detection aims at segmenting
regions corresponding to moving objects from the rest Automatic Video Based Surveillance System for
of an image. Subsequent processes such as tracking Abnormal Behavior Detection
and behavior recognition are greatly dependent on it.
The process of motion detection usually involves Automatic Video Based Surveillance System for
environment modelling, motion segmentation, and Abnormal Behavior Detection by G. H. Raisoni (2013)
object classification, which intersect with each other [34]. It presented the scenario for the proposed system
during processing. as it scanned the video input devices attached to the
system. Further, these devices were made available for
Model-based tracking algorithms track objects by selection by the user; so that the most appropriate area
matching projected object models produced with prior to be captured for a given instance of time can be
knowledge to image data. The models are usually chosen. It also included Background Image
constructed off-line with manual measurement, CAD Acquisition. This section basically consisted of
tools or computer vision techniques. As model-based capturing a frame as the reference image or the desired
rigid object tracking and model-based non rigid object ideal background condition which was considered as
tracking are quite different, we reviewed separately the reference for any further processing. Once the
model-based human body tracking (non-rigid object background was set, the camera was then programmed
tracking) and model-based vehicle tracking(rigid for capturing live video footage of the monitored area.
object tracking) [32]. Then Pre-processing was performed on the acquired
images for enhancing the quality of the frames. The
The approach of detecting object and using digital video frames had a lot of noise due to camera,
illumination, reflections and others. The process of Suspicious Object Detection Using
Back-Tracing Technique study is similar to this study.
This could be removed and quality of images was It captured videos from the camera then started
enhanced with the help of pre-processing stages. The analyzing the images provided by the camera to the
captured video was to be processed for detecting any system where the image processing came in.
change with the reference frame taken and processed Afterwards, it would start tracking and detecting the
in the previous step, as the reference condition. The object. The system would analyze the monitoring of a
Image Acquisition Toolbox available within Matlab particular time and would send it to the system for
was used, for camera controlling connected to the report and alarm. While the proposed study had the
personal computer, then it applied Foreground real time notification to the security personnel and
Extraction. This was one of the significant sections of through the server side, it would also alarm the server
the project. Here, the foreground object was side by making a sound notification to the
reconstructed by removing the background elements. administrator. A report or activity log was present on
the server side for the system administrator to be
To accomplish this, the background image obtained updated or informed.
previously was taken as a perfect reference image for
the processing of image segmentation. To achieve this, Suspicious Object Detection System
it had algorithm for change detection in case an object
was detected in an abandoned object detection. The Suspicious Object Detection System by Mrunalinee
timer was initiated and continuously incremented until Patole, Abhishek Charwad, Akshay Mengade, Naresh
found that the object was static. The incremented timer Choudhary and Vrushali Chondhe (2017). The system
was consistently compared against a pre-defined contained the following modules: 1. Background
threshold timer value. segmentation. In this module, a live feed of the camera
was processed in which if it detected any abandoned
They had similar idea or concept when it came to object, then the system would set it as a background
detecting and tracking object. It used thresholding object for further processing. 2. Change detection. A
approach to separate the background and the suspicious change was observed and noted down, then
foreground image similar to this system. The only an alert was generated and the notification was sent to
difference between the two studies was that, it could the authority. 3. Tracking. In this module, the systems
detect or identify abnormal human behavior because focused on area selected in detection and tracked the
its system had a data set of pre-defined normal human persons which were close to that abandoned object
behavior or activity. [36]. Lastly, Alarm and Display. In this module, if a
person did not come back within a specified time
Suspicious Object Detection Using Back-Tracing period then the alarm event get triggered and this
Technique showed an abandoned object was detected.
Suspicious Object Detection Using Back-Tracing The concept of background extraction and foreground
Technique by Aniket Bhondave (2016) [35]. The extraction was present in this study, which was a
scenario for the proposed system used OpenCV (Open similar approach made to accomplish this study. It
Source Computer Vision Library). It is an open source used background segmentation then after that if there
computer idea and machine learning software public was a difference between the first frame, the
library. OpenCV is manufactured to provide a background to the following frames which was the
common structure for computer image applications foreground, then it would start analyzing and tracking
and to hasten the use of machine view in the the movement. It also used an alarm notification
commercial products. Being a BSD-licensed product, towards a given time when the object was left or
OpenCV makes it easy for productions to operate and abandoned. However, it was not fully developed but
modify the code. It incorporates Java Virtual Machine the idea and concept was there. Also, it did not have
(JVM). security features or video recording. This made a
difference between the two studies.
It is customary of computer software programs and
data structures that use a virtual machine exemplary Suspicious Object Detection
for the execution of supplementary computer programs
and scripts. The model used by a JVM admitted a This paper on Suspicious Object Detection by
form of computer transitional language commonly Tarushikha Joshi (2014), a method for automatic
referred to as Java byte code. detection of a suspicious object in public places. The
suspicious object would be carried by a person in the The concept of background extraction and foreground
scene. An object area was extracted using background extraction is present in this study which is the same
subtraction. For selecting an object of interest, a approach made to accomplish the Suspicious Object
unique label had been assigned to all objects. For Detection with Alarm Notification for Security
labelling of objects, connected component labelling Personnel study. It used background segmentation.
algorithm was used. However, the system did not track the object, it only
detected suspicious activity. In this study, it has a data
A suspicious object was any object unattended for a set given for a particular activity scenario that a system
specific time period. An object could be declared as could detect. Once the system compares the current
suspicious on the basis of the distance of the object activity to the data set, it would have its results. This
from its owner for a specific time period. Matlab was idea needs more improvements in terms of accuracy
used to develop this system [37]. and efficiency.
First, it starts with counting the number of objects. If A Framework for Suspicious Object Detection from
the number of objects is equal to two then it is checked Surveillance Video
in the array of areas that the area of the object has been
decreased with respect to the previous frame or not. If This paper about A Framework for Suspicious Object
an area is decreased that means a split has occurred, Detection from Surveillance Video by Rajesh Kumar
i.e. the owner left his bag. If a split is detected, then Tripathi (2014) presents a method to detect a
the distance of that object from the person is suspicious object from surveillance video [39]. It
calculated. The distance is calculated using the consists of three main steps. In the first step,
background subtraction was performed through
distance formula. If the distance between the static
background modeling using a running average method
object and moving person was greater than the
which made the system efficient to detect foreground
specified threshold value, we would declare the object
objects. In the second step, static objects were detected
is declared as suspicious and an alarm is triggered.
from foreground frames through contour features and
geometric histogram of an object. In the third step,
The concept of tracking or detection using background
static objects were classified into human and non-
extraction approach is the same and applied with the
human objects using skin color region detection and
current study. This concept also incorporated a
edge-based object recognition method. Edge-based
notification system which is similar with the proposed
object recognition method was efficient to recognize
system. However, the analysis is different. This system
full and partially visible object. If static object is a non-
analyzed the size or the area of the object detected.
human, an alarm was raised after a specified time.
Once it splits into two objects, it could start the
calculation for distance and abandonment. Unlike with The similarity of this study to the proposed system
this system which could automatically detect that it emphasizes that the first process it performed was the
was suspicious or abandoned if it was not moving in background extraction and analysis which were both
the current location for a given time. present in the proposed system. Another similarity is
that it would track and detect objects. However, on the
Suspicious Object Detection in Surveillance Videos following procedures, it had a different approach
for Security Applications although it could detect objects and send an alarm
notification after it had a valid detection similar to the
This paper on Suspicious Object Detection in proposed system. It also used contour features and
Surveillance Videos For Security Applications by geometric histograms for detecting objects. Another
Trupti M. Pandit (2015) presents a system for video difference between them is that could compare or
surveillance - based applications that automatically identify human to non-human objects which was not
detected the suspicious activity [38]. Static object present in the proposed study. But, this idea can also
detection is an initial step in video surveillance. There be implemented in the proposed system for further
are enormous algorithms used for detecting static improvement and more effective approach in detecting
objects. The input video is taken from the static objects.
camera. The main strength of the approach lied in the
ability to separate background and foreground in
Methodology
accurate way from the video. The speed of the
proposed system was directly proportional to the
HD/Full HD/4K surveillance system. The study was developed using a prototyping software
development methodology. Software prototyping is the From the video input, the system performed a
activity of creating prototypes of software background extraction for the first frame of the live
applications. They are incomplete versions of the feed then it would do the foreground image acquisition
software program being developed. The prototyping to compare the first frame to the updated foreground
methodology has several benefits. It is a part of data. After which, it would perform the comparison
software development wherein, the software designer with noise and thresholding algorithm for a smoother
and implementer could get valuable feedback from the and more precise data presentation or analysis.
users early in the project. The client and the contractor
could compare if the software made matched with the If a non-moving object was detected, it would start
software specifications according to which the applying the contour to the detected objects in the
software program was built. It also allowed the foreground for tracking purposes.
software initial project estimates and whether the
deadlines and milestones proposed could be The administrator managed the accounts and viewed
successfully met [40]. Figure 1 illustrates the the recordings, notifications, and logged-ins while the
prototyping software life cycle model. security personnel would receive a notification once
the system detected a non-moving object from the
surveillance camera. The main server would be
notified as well if the system detected a static object in
the camera.
The requirement analysis showed the minimum The major solution for the system was entirely built
requirements of the hardware and software needed in using python language since python was an
the development and implementation of the system. interpreted, high-level, general-purpose programming
language. The system was fully developed using this
Table 1 shows the minimum hardware requirements of language since it had tons of libraries. With the help of
the system. opencv libraries, it expanded and incorporated more
algorithm available for the system. The minimum
Table 1. Hardware Specification requirements for python software was at least Intel
Atom processor, Disk space of 1GB, Operating
System of Windows 7 or later.
Twi lio a l l o w e d s o f t wa r e d e v e l o p e r s to
programmatically make and receive phone calls, send
and receive text messages and perform other
communication functions using its web service APIs.
Figure 8 shows the successful text message from twilio
services.
System Notification
Table 3 shows the summary of the evaluation results about the over the network. (2) The camera could detect and
system’s quality in terms of ISO 25010 standards. track suspicious objects based on the excellent
evaluation of its performance efficiency and reliability.
The tracking and detection were dependent on the
position and angle of the camera which could be
affected by poor weather condition. Although it could
track and detect objects, it had no capability to identify
the objects detected. (3) The system could send text
notifications and alarms once a suspicious object was
detected and could be accessed on any type of mobile
phones. It had an excellent evaluation result in its
portability. The system could identify which camera
had detected an object. However, the location
contained only the numeric code that needed to be
Results shows that the IT experts’ overall evaluation
memorized to identify which area it was located. (4)
on the software quality characteristic of the system
The over-all result of the system’s evaluation based on
based on ISO 25010 criteria was excellent with an
the ISO 25010 criteria was excellent.
overall mean of 4.73. Specifically the rating for
functional suitability was 4.63, performance efficiency The following conclusions were drawn after the
was 4.73, compatibility was 4.80, usability was 4.78, conduct of the study: (1) The quality of the camera
reliability was 4.73, security was 4.79, maintainability contribute to the functional suitability of the system
was 4.66, and portability was 4.73. The IT experts and connectivity via the network would improve its
commended the performance and the functionality of accessibility. (2) The tracking and detection of objects
the system that it was capable of tracking and is dependent on the positioning of the camera.
detecting objects. Although the overall rating was Increasing the number of cameras would provide a
excellent, they highlighted that it could be upgraded by more efficient tracking. (3) The system is capable of
adding more complex and secure algorithms to identifying the camera which detected the object and
improve the system. identified the location using the numeric code.
Converting the numeric code into an actual name of
The results show that the security personnel’s overall
the location would provide easy identification on the
evaluation on the software quality characteristic of the
location of the suspicious object. (4) The system
system based on ISO 25010 criteria was also excellent
passed the ISO quality standards for software
with an overall mean of 4.78. The security personnel’s
development.
rating on functional suitability was 4.70, performance
efficiency was 4.77, compatibility was 4.85, usability The following recommendations have been forwarded:
was 4.82, reliability was 4.80, security was 4.80, (1) An IP Camera was recommended to be used and to
maintainability was 4.72 and portability was 4.77. be connected over the network for faster access. (2)
The system should be tested using multiple cameras
The evaluation results in all aspects of the software
located in different locations to improve the tracking
quality characteristic rated by the security personnel
and detection processes. A high definition camera was
were much higher compared to the IT experts. The
recommended to efficiently work in a poor weather
evaluators put emphasis and appreciation on the
condition. Moreover, a future study using machine
compatibility, reliability, security and most especially
learning may be conducted to add features of object
on the usability of the system. Since it was secured and
identification. (3) Numeric location code should be
easy to use, the security personnel evaluators agreed
converted into the actual location name of the area for
that it was a useful system for them.
easy identification. (4) Lastly, it was recommended
that the system may be implemented to establishments
Conclusion due to its excellent performance and reliable features.
domain experts. It was attributed to the low quality of [2] Ball State University. Suspicious Object.
the plug and play camera which was not connected
[3] Best Coursera Courses for Security & Safety Challenges in a Different Types Of Images Different Types Of Images Different
Globalized World - Best Coursera Course Finder. [Online]. Types Of Images Different Types Of Images. International Journal
of Computer Science Issues. 7.
[4] Rechenberg and A. Rechenberg, “6 Technological Developments
in CCTV Surveillance Systems,” Rechenberg, 28-May-2019. [25] D.S. Guru, Y.H. Sharath Kumar, S. Manjunath, Textural
[Online]. features in flower classification, Mathematical and Computer
Modelling, Volume 54, Issues 3–4, 2011, Pages 1030-1036, ISSN
[5] T. A. Scally, “State of the Market: Video Surveillance 2018,” 0895-7177,
SDM Magazine RSS, 01-Mar-2018. [Online].
[26] “Image Thresholding¶,” OpenCV. [Online].
[6] S. R. J. Ramson, A. Bashar, J. S. Raj, and Ditzinger (2020).
Innovative Data Communication Technologies and Application. [27] “Otsu's method ,” Everything Explained Today.
Springer International Publishing, 2020.
[28] Tsai, C. M., & Lee, H. J. (2002). Binarization of color
[7] P. Barter, “On-Street Parking Management,” Federal Ministry document images via luminance and saturation color features. IEEE
for Economic Cooperation and Development, 12-Jan-2020. Transactions on Image Processing, 11(4), 434-451
[8] D. M. West and J. R. Allen, “How artificial intelligence is [29] Jiao, L., Zhang, F., Liu, F., Yang, S., Li, L., Feng, Z., & Qu, R.
transforming the world,” Brookings, 28-Apr-2020. [Online]. (2019). A Survey of Deep Learning-Based Object Detection. IEEE
Access, 7, 128837-128868.
[9] “ALARM NOTIFICATION,” Cambridge Dictionary.
[30] D. Adalinge1 and D. Patil, “Survey on Abandoned Object
[Accessed: 12-March-2020]. Detection using Image Processing,” International Journal of
Innovative Research in Computer and Communication Engineering,
[10] “Camera,” Merriam-Webster. vol. 5, no. 5, May 2007.
[11] “Image,” LALI. [Online]. [31] Danker, A. J., & Rosenfeld, A. (1981). Blob detection by
relaxation. IEEE Transactions on Pattern Analysis and Machine
[12] G. B. Shelly and M. Vermaat, Discovering computers: Intelligence, (1), 79-92.
fundamentals: your interactive guide to the digital world. Boston,
MA: Course Technology, 2012. [32] Mishra, M. S. K., Jtmcoe, F., & Bhagat, K. S. (2015). A survey
on human motion detection and surveillance. International Journal of
[13] Jiao, L., Zhang, F., Liu, F., Yang, S., Li, L., Feng, Z., & Qu, R. Advanced Research in Electronics and Communication Engineering
(2019). A Survey of Deep Learning-Based Object Detection. IEEE (IJARECE) Volume, 4.
Access, 7, 128837-128868.
[33] Jung, S., Cho, Y., Kim, D., & Chang, M. (2020). Moving object
[14] The Editors of Encyclopedia Britannica, “Security Alarm,” detection from moving camera image sequences using an inertial
Encyclopedia Britannica, 17-Aug-2012. [Online]. measurement unit sensor. Applied Sciences, 10(1), 268.
[15] “Security Personnel,” Security personnel | legal definition of [34] Lavanya, M. P. (2014). Real time motion detection using
Security personnel by Law Insider. [Online]. background subtraction method and frame difference. Int. J. Sci.
Res.(IJSR), 3(6), 1857-1861.
[16] D. Clark, “What is a Suspicious Item?: Suspicious Packages,”
Engage in Learning, 11-Feb-2020. [Online]. [35] Gary Bradski and Adrian Kaehler. Learning OpenCV:
Computer vision with the OpenCV library. " O’Reilly Media, Inc.",
[17] “TRACK” Cambridge English Dictionary.
2008.
[Accessed: 3-April-2020].
[36] Luna, E., San Miguel, J. C., Ortego, D., & Martínez, J. M.
(2018). Abandoned object detection in video-surveillance: survey
[18] “Video Camera,” pcmag.com.
and comparison. Sensors, 18(12), 4290.
[19] W. Wei, Building service-oriented government: lessons,
[37] Joshi, T., Aarya, H., & Kumar, P. (2016). Suspicious object
challenges and prospects. Singapore: World Scientific, 2013.
detection. In 2016 2nd International Conference on Advances in
[20] “SECURITY CAMERA: TECHNOLOGY GATE,” Computing, Communication, & Automation (ICACCA)(Fall) (pp. 1-
SECURITY CAMERA | TECHNOLOGY GATE. 6). IEEE.
[21] “Deep Machine Learning AI,” Deep Machine Learning AI. [38] Paul, M., Haque, S. M., & Chakraborty, S. (2013). Human
detection in surveillance videos and its applications-a review.
[22] Grigorova, K., Malysheva, E., & Bobrovskiy, S. (2017). EURASIP Journal on Advances in Signal Processing, 2013(1), 176.
Application of Data Mining and Process Mining approaches for
improving e-Learning Processes. In Информационные технологии [39] Tripathi, R.K., Jalal, A.S. & Agrawal, S.C. Suspicious human
и нанотехнологии (ИТНТ-2017) (pp. 1960-1966). activity recognition: a review. Artif Intell Rev 50, 283–339 (2018).
[24] Singh, Krishna Kant & Singh, Akansha. (2010). A Study Of Benjamin L. Cornelio Jr.
Image Segmentation Algorithms For A Study Of Image
Segmentation Algorithms For A Study Of Image Segmentation
West Visayas State University, Janiuay Campus
Algorithms For A Study Of Image Segmentation Algorithms For Iloilo City, Philippines