Hidden Object Detection for Computer Vision Based Test Automation System

Hidden Object Detection for Computer Vision Based Test Automation
System
RAJIBUL ANAM, MD. SYEFUL ISLAM, MOHAMMAD OBAIDUL HAQUE
Samsung R&D Institute Bangladesh
BANGLADESH
rajibul.an@samsung.com, syeful.islam@samsung.com, obaidul.h@samsung.com
Abstract: In Software Quality Assurance, computer vision based automation tools are used to test the window
based application and window contains many type of objects like button, box, list, etc. Automation tool detect
window objects by comparing images. Most of the objects are visible in the screen but some are not visible to
the screen at the first time, proper interaction with the window application hidden objects get visible to the
screen like drop-down list item, editor text object, list box item and slider. With the vision based automation
systems these hidden objects cannot be searched directly. In this paper proposes some methods which use
image and shortcut key to interact with the testing software to search the hidden objects. These methods will
enhance the automation tools to access the window application hidden objects faster.
Key-Words: - test case automation; software quality assurance; vison based; window application
1 Introduction
Software Quality Assurance (QA) is one of the
critical areas of software development process life
cycle. After co-work with developer and designers,
QA ensures the correctness of the operation by
testing the software application through different
type of test cases [1]. Many methods have been used
to test the software and among them Black Box and
White Box Testing are very commonly used. Black
Box testing consists of specification and experience
based testing, which checks the entire software
operation [1-2]. White Box testing follows the
structure based testing, which checks the software
process flow [1-2]. QA testing actions or steps are
executed by mouse and key events, after the events,
program flows and interfaces get change [2], which
is the part of the QA testing. Manual testing
operates by human, it executes series of steps and
check for the specific output which has chances of
error [1],[3]. But Automation System executes
series of steps according to the code instruction,
which executes test steps faster than human and less
error [3]. Moreover, automation system has been
used for Black Box testing because it follows
specific test steps and expects for target results.
Most of the testing application needs to be tested
with the predefined Test Cases (TC). These
predefined TCs can be automated, so the testing will
be faster and human dependency will be reduced.
Graphical User Interface (GUI) QA testing
purpose many type of automation system have been
used such as Pesto [4], DEVSimPy [5], Watir [1],
Selenium [7], Sikuli [6],[8]. These systems either
use the vision based or screen objects position
pointing technique to track the screen objects. Script
re-usability and smooth execution are essential for
the automation system [6]. Automation system
executes QA testing steps/actions easily by tracing
image and objects position. GUI applications has
hidden or not visible objects like text objects in the
editor, drop-down list object, multi tab scroll object
and slider. These types of hidden objects can be
searched easily by manual QA testing. But vision
based systems firstly; trace objects (image) and
secondly, execute action events on the screen object
position. Vision based system uses only image
based object detection method, which uses only
mouse events to interact with the GUI testing
application therefore, complex steps like search
hidden object from a list box or slider scrolling
takes time and sometimes failed to trace the target
hidden object. Considering these difficulties,
focuses on how to access hidden objects accurately
and enhance re-usability.
This paper proposes some methods that will
enhance vision based automation tools to discover
hidden objects (GUI elements) from the GUI based
applications by using key and mouse events. Vision
based system uses only mouse events therefore,
include the key events (shortcut key additional
feature) to trace the hidden object easily and
accurately. The propose methods use shortcut key
WSEAS TRANSACTIONS on COMPUTERS Rajibul Anam, Md. Syeful Islam, Mohammad Obaidul Haque
E-ISSN: 2224-2872 570 Volume 14, 2015

and click (image object to click on the GUI
application) for action events, then interact with the
visible objects through few steps, afterwards the
hidden object gets visible on the screen. The
proposed methods will enhance the vision based
automation systems to search the hidden objects
faster.
This paper is arranged as follows. Section 2
provides brief review of other automation tools for
GUI testing. Section 3 describes about the proposed
solution. Section 4 details of the usability study and
finally conclusion is on section 5.
2 Related Works
Software QA testing automation systems has been
used to reduce human effort. Automation systems
trace the target objects by screen object position,
image matching and source name. Automation
actions/steps are executes by click, drag-and-drop
and keyboard events. Tools like Sikuli, Robot and
Pesto uses image tracing, object source name and
object screen position to access the object. These
automation systems developed either on vision
based or screen objects position detection based
methods. Below discuss details of these two types of
automation systems.
2.1 Vision Based Automation
Sikuli is an open source GUI vision based
automation system, which searches the target object
using screenshot [8-10]. The IDE permits users to
take a screenshot of the target object (GUI elements)
such as button, icon, dialog box and run time detects
the object to direct the mouse and keyboard events
[11-13]. Figure 1 illustrates the Sikuli Framework,
where built-in modules are available like find, click
and key events [14]. There are more modules
available which cannot be used from IDE directly. It
has the Application Programming Interface (API)
for testing and developing the library. It is a
platform independent framework.
Robot Framework is a generic testing automation
system to test Acceptance Test-Driven Development
(ATDD) [15]. ATDD is a process where developers
and testers discuss the demands required by the
customers to come with the acceptance test before
development. The acceptance test provides the
functional importance of the software [15].
Figure 1. Sikuli Framework
Figure 2. Robot Framework
The aim of the acceptance tests is to justify the
requirements by providing examples for each test.
The examples can be tested to prove compliance.
The script language is written using plain English
natural commands called Keywords [9]. Keywords
are common like methods in programming
language. Natural command keywords make the
tests more readable and easy to understand even for
non-coders. This framework script writing is
extended to Python (can run also on both Ironpython
and Jython) or java. The developers can use the
existing syntax to create the script or can create own
syntax. Robot framework uses for GUI testing and
system resource management, but only java based
software can be tested. It generates auto report of
the testing as html and text format. It has the API for
testing and developing the library.
2.2 Screen Objects Position Detection Based
Automation System
The TestComplete [16], TestPlant [17] and Squish
[18] are recoding based framework. These systems
E-ISSN: 2224-2872 571 Volume 14, 2015

record user interaction events according to the
screen object position and replay time these systems
execute the events according to the recorded
sequence. Replay time these systems execute the TC
very fast and generate a report of the TC. These
systems use their own language to generate script
and developers can edit the auto generated scripts.
Recording time these systems track the user events
position like mouse click position, drag-and-drop
positions and keyboard input information. If the
testing software font, color gets change then the
recorded script can be replay unless the application
layout has changes. Moreover recording and replay
(execution) time the screen size (resolution) should
be same because all the action events execute
according to the objects position. If the app window
or pop-up window position gets change on the
screen then replay time generate errors. The
application layout should be same at recording and
replay time.
3 Proposed Solution
Various types of automation systems are available
and most of it uses vision based algorithm to search
the GUI target objects. These systems take the
screenshot of the window first; then select the target
object from the screenshot and interact with the
application by mouse or key events. These systems
are used to search the GUI window target objects
like toolbar button, menu item, icon and dialog box
[14]. Moreover, searching arbitrary depends on the
screenshot and the target object image. If the target
objects image do not matches with the screenshot
image, then automation system could not search the
target object in the screen. In this case system
searches for an object which is not visible (hidden
object) in the screen or the object is not available in
the testing software. But the hidden object needs to
be search because it is available in the testing
software and part of the TC. The automation system
would not be able to search the target object until
the target object gets visible on the screen. Current
approaches required entering an image as query to
search the target object. If searching for a hidden
object in the window screen and could not trace the
target object then it will generate an error, which is
a limitation of the vision based automation system.
The proposed method searches hidden objects
like item in the editor, drop-down list object, multi
tab scroll object and slider positions. In addition
shortcut key events has introduced instead of
screens object image to trace the target object. The
Editor Scroll-bar Object Selection method uses to
search the hidden object from a scroll-bar affiliated
object. The Drop-down List Object Selection
method is applicable to search hidden object from
the drop-down list. The Multi Tab List Object
Selection method is valid to search the hidden object
from the multi tab list box and Slider Positon
Selection method is applicable to search and puts
the slider position according to the code instruction.
Below sections discuss details of the proposed
methods which uses mouse and key (shortcut key)
events.
3.1 Editor Scroll-bar Object Selection
(ESOS)
QA testing time automation system needs to check
the text editor or webpage interface and font
decoration objects. It becomes very hard to find text
object in the editor which contains a long page and
the target text object stand at the end of the page, at
this scenario the stroller get enabled. But hidden
objects do not appear on the screen and could be
visible unless the system searches for the hidden
objects [14]. Figure 3 shows the editor screen with
scroll-bar object, where a text editor is opened. The
automation system needs to search the figure 4
hidden target object (search in the screen) in the
(figure 3) text editor. In this scenario the automation
system needs to scroll down the scroll-bar using the
mouse [13]. There is no specific method for the
scroll-bar to scroll down at specific point. To solve
this problem, proposed the ESOS method where the
scroll-bar will be scroll down until it reaches the
target object.
Figure 5 line 1-2 searches the mainobject (the
style.css object) and take focus on the style.css
object. Line 3 puts the cursor at the beginning of the
editor. Line 4-8 searches for the targetobject (figure
4), if it does not find targetobject, then goes to the
next line until it reaches to the targetobject (figure
4). If the targetobject found then select (click) the
hidden target object. The scroll-bar cannot be used
directly (can access it but cannot scroll it as
requires) and with this method automation system
can search the hidden object from the screen without
scrolling the scroll-bar. Figure 6 shows where the
hidden object gets visible and hidden target object is
found by using this method.
E-ISSN: 2224-2872 572 Volume 14, 2015

Figure 3. Screenshots of a Scrolling Object
Figure 4. Hidden Target Object
Figure 5. Editor Scrollbar Object Selection Method
Figure 6. After Scrolling the Screen
3.2 Drop-down List Object Selection
(DLOS)
Drop-down list objects are used in the window and
web based applications. There is some drop-down
list which contains text with images. Most of the
drop-down list contains long item lists which need
to check and test for the QA. If drop-down item list
is long, then most of the items will not be visible on
the screen [14] which creates hidden object in the
list. But drop-down list hidden object items cannot
be access properly by the automation system
because the scroll-bar appears dynamically and
needs to scroll it to get the hidden object. The
proposed DLOS method will enhance the
automation system to search the hidden objects from
the drop-down list.
Figure 7. Screenshot of Dropdown List
Figure 9. Dropdown List Object Selection Method
Figure 7 shows the drop-down list with long item
lists which contains hidden object. Figure 8 shows
the hidden target object which needs to search from
the drop-down list. Figure 9 shows the DLOS
algorithm where line 1 uses shortcutkey to select the
drop-down list object. Line 2-6 searches for the
targetobject from the list, if not found targetobject
then goes to the next list item until it reaches to the
targetobject. Figure 10 shows the targetobject
style.css object
Scrollbar needs to scroll down
Scrollbar needs to scroll down
Input:
mainobject is an image or text object to focus on the object;
targetobject is an image object;
Output:
targetobject get selected;
Variables:
screenimage is the desktop screen capture image;
onelinedown is an keyboard value to move down the cursor
next line;
ScrollbarObjectSelection(mainobject, targetobject)
1. If mainobject matched with screenimage Then
2. Click on the mainobject;
3. Put cursor to the beginning of the editor;
4. While until found the targerobject
5. Move the cursor onelinedown;
6. If targetobject matched with the screenimage Then
7. Click the targetobject in the screen;
8. Break;
9. Else cannot found the targetobject;
10. End
Input:
shortcutkey is keyboard value to take focus of the object;
Output:
Variables:
onelinedown is an keyboard value to move down the cursor
next line;
DropdownListObjectSelection(shortcutkey, targetobject)
1. If shortcutkey works to select the object Then
2. While until the targetobject
4. If targetobject matched with screenimage Then
6. Break;
8. End
E-ISSN: 2224-2872 573 Volume 14, 2015

matched with screenimage. This method searches
the entire hidden objects from the list and checks for
the target object.
Figure 10. Screenshot of the Dropdown List with Target Object
3.3 Multi Tab List Object Selection
(MTLOS)
QA testing time needs to interact with GUI window
multi tab objects [13]. A multi tab window contains
more than one list box objects with scroll-bar
features. List box object contain many hidden
objects, automation system needs to interact with
the hidden objects to complete the TC. Figure 11
shows an example of the multi tab objects which
contains three tab objects. Figure 11 (1) shows the
first tab object (sample), figure 11 (2) shows the
second tab object (Web App) with list box, figure 11
(3) shows the another tab object with hidden (item
list) and figure 11 (4) scroll-bar enabled for
scrolling.
Figure 11. Multi Tab List Object Screenshot
To search the target object (figure 12) from figure
11, firstly needs to select the (1) sample object,
secondly select the (2) web app object from the list
box and finally searches for the target object from
the (3) next list box.
The proposed MTLOS method is able to search
target object from multi tab list-box hidden object.
Figure 13 shows the MTLOS method, where line 1-
2 selects the mainobject (figure 11, object 1), line 3-
4 selects the next tab object and line 5-9 searches
(figure 11, object 2) for the firstkeyinfo image
object until it found. Line 10-11 selects next tab
object (figure 11, object 3), line 12-16 search for the
tergetobject (figure 12) until it found. Figure 14
shows the target object found using this method.
Figure 13. Multi Tab List Object Selection Method
Input:
mainobject is an image object;
firsttab is a keyboard tab value;
firstkeyinfo is an image object;
secondtab is a keyboard tab value;
Output:
Variables:
onelinedown is an action variable to move down the cursor
next line;
Multitabobjectselection(mainobject, firsttab, firstkeyinfo, secondtab,
targetobject)
3. If firsttab is true Then
4. Press tabkey;
5. While until firstkeyinfo
7. If firstkeyinfo matched with screenimage Then
8. Click the firstkeyinfo in the screen;
9. Break;
10. If secondtab is true Then
11. Press tabkey;
12. While until targetobject
16. Break;
18. End
1
2
3
4
E-ISSN: 2224-2872 574 Volume 14, 2015

Figure 14. Screenshot of Multi Tab List Object with Target Object
3.4 Slider Position Selection (SPS)
Slider is a GUI window application object which
has no onscreen values (from where to drag and
drop) like scroll-bar. QA purpose slider needs to
access, change slider positions and checks the
expected result. But for the QA testing purpose
automation tool needs to access and change the
value of the slider [13] which takes time and
sometimes automation system failed to change the
slider positions as requires. Figure 15 shows the
slider where it is at the Error mode and testing
purpose needs to set as Debug mode (figure 16).
While TC execution, automation system cannot put
the slider as required position easily, sometimes it
starts scrolling on the left and sometimes on the
right side. As a result it takes additional time to
reach to the goal. The proposed SPS method can
overcome this problem and put the slider position
according to the code instruction easily.
Figure 15. Screenshots of Slider Object
Figure 16. Target Object
Figure 17. Slider Positon Selection Method
Figure 18. Slider Position with Minimum Value
Figure 17 illustrates the SPS methods, where line
1-2 take focus on the sliderobject, line 3 takes the
slider lowest sliding position (figure 18). Line 4-8
increases the slider positions according to the
keyinfo value until it reaches to the targetobject.
With this method the slider hidden objects value can
be search faster.
3.5 Framework Comparisons
The Sikuli [9] is a vision based automation
framework, it uses image to detect the object and
after that use the action events to interact with the
testing software. The TestComplete [16], TestPlant
[17] and Squish [18] are screen objects position
detection based framework, all the mouse and key
events are applied on specific point of window
screen object position. If the object position gets
change (screen resolution) or mismatched then
select or interact with different objects and generate
error. Table 1 shows the comparison criteria of
automation frameworks.
Screen objects position detection based
framework can access the hidden objects, if the
screen window objects position remains fixed on the
second run time. But TC execution time it is very
hard to confirm the window objects position.
Developer record the test, execution time if the
window object appears at different position, then
developer needs record the steps again which is
Input:
mainobject is an image object;
keyinfo is a keyboard key move value;
Output:
Variables:
onelinedown is an action variable to move down the cursor
next line;
scrollslider(mainobject, keyinfo, targetobject)
3. Scroll slider to lowest value;
4. While until targetobject
5. Press keyinfo;
7. targetobject object found;
8. Break;
10. End
E-ISSN: 2224-2872 575 Volume 14, 2015

redundant. Table 2 illustrates the frameworks, if the
screen window object position or the resolution gets
change, then cannot interact with the onscreen
window objects to search the hidden objects. But
same time the proposed methods are able to interact
with the changed screen (position or resolution)
window objects. The proposed methods uses key
and (image objects) click events to interact with the
window objects to search the target hidden objects.
Table 1. Comparison of Automation Framework Criteria
Sikuli TestComplete TestPlant Squish
Open Source Yes No No No
App code
required
Yes Yes Yes Yes
Platform
Independent
Yes No Yes Yes
Hidden
Object
Identification
No No Yes Yes
Image Based Yes No No No
Screen
Position
Dependent
No Yes Yes Yes
Test
Recording
No Yes Yes Yes
Table 2. Comparison of Framework With Screen Object Position
Criteria
Sikuli
TestComplete
TestPlant
Squish
Propose
Methods
Changed
Screen
Position
Hidden Object
Identification
No No No No Yes
Slider Interaction
No No No No Yes
3.6 Complexity Comparisons
The time complexity depends on flow of the
algorithm [19]. If the algorithm uses nested
operation then the complexity gets higher. Below
table 3 shows the comparison of the Proposed
Algorithm (PA) and existing Vision Based
Algorithm (VA), where O denotes as growth of a
function and n is number of steps. It is clear that VA
and PA time complexity are almost same. There is
no significant difference between PA (ESOS,
DLOS, MTLOS and SPS) and VA. But there are
differences on the execution time because of the
dependency (wait for the object, interaction
methods).
Table 3. Complexity of the Algorithms
Time Complexity of VA Time Complexity PA
O(n) [14] O(n) ESOS
O(n) [14] O(n) DLOS
O(n) [14] O(n) MTLOS
O(n) [14] O(n) SPS
4 Usability Study
GUI Automation system executes action according
to the instructions (code). Basically two types of
event occur in the GUI automation, one is key event
and another is mouse event. Automation system
runs the code; execute commands which interact
with GUI testing system. To generate mouse or key
events, screen objects position detection based
systems record the user actions and automatic
generates code for automation system. And image
based systems do not have this facility, developer
needs to write code.
4.1 Case Study Design
This section describes the experimental results
obtained by the VA and PA with four predefined
Test Cases. The empirical study presented in this
paper is conducted in real time context. This paper
proposed four methods which uses vision based
methods and shortcut key to access the object,
which is a combination of image and key events.
Moreover these methods will enhance the
automation system to get the target object faster. To
support this claim carried out a case study to test the
hypothesis below.
H1: Using shortcut key (key event) and vision
based screen object detection (click) to search the
target object reduces automation systems interaction
events than using only vision based object detection
(click).
H2: Combination of vision based screen object
detection and shortcut key can trace the hidden
object faster than using only vision based screen
object detection.
This study was designed to test the VA and PA
performances. To execute the automation used Intel
Core i7 (3.4GHz) processor, 4GB ram, Windows 7
OS and display resolution (1920×1080). QA testing
purpose selects the Tizen IDE application [20].
Table 4 shows details of TCs, which is created to
test the Tizen IDE for QA purpose. Each TC was
executed thirty times randomly. TC-1 executes the
ESOS, TC-2 executes the DLOS, TC-3 executes the
MTLOS and TC-4 executes the SPS algorithm.
There are two dependent variables in this study:
number of events (interaction) and task completion
E-ISSN: 2224-2872 576 Volume 14, 2015

time. This information cannot observe directly and
therefore, can only be measured after completion of
all the data.
Table 4. Test Cases
Test
Case
Steps Search the
Target
Object
Proposed
Method
Used
TC-1
Click on style.css file from
the file browser
footer p {
font-size:
12px;}
ESOS
TC-2
Click Top-up menu Window-
>Preferences; New window
Double Click Web->Click
CSS Files->Encoding List
Korean, ISO
2022
DLOS
TC-3
Click Top-up menu File-
>new->Tizen Web Project;
New window Click Sample-
>Mobile->Web App-
>TizenWinset
TizenWinset MTLOS
TC-4
Click Top-up menu Window-
>Preferences; New window
click Tizen SDK->Click
Logging->Slider
Select Slider
to DEBUG
SPS
4.2 Results
Figure 19 shows the number of mouse and key
events has been used to execute the TCs. SA used
52 clicks; ESOS used 2 clicks and 2 key events to
execute TC-1. SA used 30 clicks; DLOS used 1
click and 8 key events to execute TC-2. To Executes
TC-3, SA used 18 clicks and MTLOS used 2 clicks
and 8 key events. SA used 13 clicks; SPS used 1
click and 7 key events to execute TC-4. The
interaction events shows that the proposed
algorithms used less interaction events compared to
only vision based systems. Figure 20 shows the total
number of events (click and key) to execute the PA
and VA. PA used six click events and 25 key events,
same time VA used 113 click events to execute all
the TCs. Table 5 illustrates the Mann-Whitney U
test analysis results, where Z = 2.411 and p =
0.0163, which is statistically significant PA used
less interaction events than VA. From this result can
conclude that combination of mouse and key events
required less interaction to execute the TCs which
supports the H1 hypothesis.
Figure 21 shows TC-1 completion time of VA
and ESOS, VA has two outlier values, VA took
30.74 seconds and ESOS took 1.91 seconds to
complete the tasks, where N = 60, x̅ of VA is 45.50,
ESOS is 15.50, Z = 6.663 and p = 0.000002, which
is statistically significant and the result shows that
ESOS takes less time that VA.
Figure 19. Number of Events for Each TC
Figure 20. Total Number of Events
Table 5. Mann-Whitney U test analysis of number of events to finish
tasks
Method Events N Mean
VA
Click 4 6.5
Key 4 2.5
PA
Click 4 2.5
Key 4 6.5
Test Statics
Z 2.411
p value (2 tailed) 0.0163
Figure 22 illustrates the TC-2 completion time,
VA has three outlier values, VA took 7.29 seconds
and DLOS took 2.96 seconds, where N=60, x̅ of SA
is 45.50 and ESOS is 15.50, Z = 6.663 and p =
0.000002, which is statistically significant and the
result shows that DLOS takes less time than VA.
Figure 23 illustrates the TC-3 completion time,
VA has five outlier values and MTLOS has one
outlier value. VA took 19.14 seconds and MTLOS
took 13.29 seconds, where N = 60, x̅ of SA is 45.50,
ESOS is 15.50, Z = 6.654 and p = 0.0000002, which
is statistically significant and the data shows that
MTLOS takes less time than VA.
E-ISSN: 2224-2872 577 Volume 14, 2015

Figure 21. TC-1 Completion Time
Figure 24 shows the TC-4 completion time, VA
has six and SPS has one outlier value. VA took
15.94 seconds and SPS took 9.85 seconds, where N
= 60, x̅ of SA is 43.50, ESOS is 17.50, Z = 5.767 and
p = 0.000008, which is statistically significant and
the results shows that SPS takes less time than VA.
Figure 25 shows the completion time of four TCs,
VA took 18.27 seconds and PA took 7 seconds.
Table 6 shows the average TCs completion time and
table 7 shows the Mann-Whitney U test analysis
results, where N = 240, x̅ of SA is 164.50, PA is
76.50, Z = 9.819 and p = 0.0000001, which is
statistically significant and the results shows that PA
takes less time than VA. From this analysis can
conclude that combination of mouse and key events,
automation systems can trace the hidden objects
faster which supports the H2 hypothesis.
Figure 24. Total TC Completion Time
TABLE 6. Average execution time
TABLE 7. MANN-WHITNEY U TEST ANALYSYS OF COMPLETION TIME TO
FINSH TASKS
Methods N
VA 120
PA 120
Total 240
Test Statistics
Z 9.819
p value (2 tailed) 0.0000001
VA Proposed Methods
Test Case Execution Time (sec) Execution Time (sec)
TC-1 30.74 1.91
TC-2 7.29 2.96
TC-3 19.14 13.29
TC-4 15.94 9.85
TOTAL (AVG) 18.27 7
E-ISSN: 2224-2872 578 Volume 14, 2015

5 Conclusion
GUI automation tools enhance Test Case execution
and reduce human efforts. Most of the Black Box
Test Cases can be executed with this system;
limitation of the technology vision based automation
system in some cases take time and failed to
find/search hidden objects and dynamic appearance
of the objects. As a result all type of TCs cannot be
executed using VA system. The proposed
techniques have the unique features to identify
hidden objects even the window objects screen
position gets change. The proposed methods are
implemented in real time automation application,
which can discover the hidden objects smoothly and
enhance the re-usability. The usability study results
show that combination of key and mouse events in
the VA system can find the hidden target object
faster. These methods enhance the VA systems to
find the target object faster which will help the QA
testers to get the result quicker. Currently there is
one limitation with these methods. It takes time to
check the list box objects one by one to search the
hidden target object. Future plan is to overcome
these two limitations and works for complete
introducing full testing framework for hidden object
detection.
References
[1] S. Inderjeet, and T. Bindia, "Comparative
Analysis of Open Source Automated Software
Testing Tools: Selenium, Sikuli and Watir",
International Journal of Information &
Computation Technology, vol. 4, pp. 1507-
1518, 2014.
[2] K. Dea-Kwang, and L. Lee-Sub, "Reverse
Engineering from Exploratory Testing to
Specification-based Testing", International
Journal of Software Engineering and Its
Applications, vol. 8(11), pp. 197-208, 2014.
[3] L. Maurizio, S. Andrea, R. Filippo, and T.
Paolo, "Automated Generation of Visual Web
Tests from DOM-based Web Tests",
ACM/SIGAPP Symposium on Applied
Computing, April, 2015.
[4] M. Leotta, A. Stocco, F. Ricca, P. Tonella,
"PESTO: A Tool for Migrating DOM-Based to
Visual Web Tests", ACM/SIGAPPSymposium
on Applied Computing, April, 2015.
[5] L. Capocchi, J.F. Santucci, T. Ville, "Software
Test Automation using DEVSimPy
Environment", International Conference on
Principles of Advanced Discrete Simulation,
May, 2013.
[6] Borjesson, and F. Robert, "Automated System
Testing using Visual GUI Testing Tools: A
Comparative Study in Industry",
"Borjesson2012visual", 2012.
[7] J. Hyunjun, L. Sukhoon, B. Doo-Kwon, "An
Image Comparing-based GUI Software Testing
Automation System", World Congress in
Computer Science, Computer Engineering, and
Applied Computing, 2012.
[8] K. Pragya, "Ameliorating the image matching
algorithm of Sikuli using Artificial Neural
Networks", International Journal of Computer
Science & Communication, vol. 5, pp. 1-4,
2014.
[9] http://www.sikulix.com
[10] C. Tsung-Hsiang, "Using graphical
representation of user interfaces as visual
references", The 24th annual ACM symposium
adjunct on User interface software and
technology, pp. 27-30, 2011.
[11] S. L. M. Jeffrey, "User interface computation
as a contextualized approach for introductory
computing instruction", The 9th Annual
International ACM Conference on International
Computing Education Research, pp. 179-180,
2013.
[12] V. Andriychenko, L. Ying-dar , C. T. National,
"Automatic Functionality and Stability Testing
Through GUI of Handheld Devices",
CiteSeerx, 2011.
[13] C. Tsung-Hsiang, Y. Tom, C. M. Robert, "GUI
testing using computer vision", CHI 10th
Conference on Human Factors in Computing
Systems, pp. 1535-1544, 2010.
[14] Y. Tom, C. Tsung-Hsiang, C. M. Robert,
"Sikuli: using GUI screenshots for search and
automation", The 22nd annual ACM
symposium on User interface software and
technology, pp. 183-192, 2009.
[15] http://robotframework, Jan, 2015.
[16] http://smartbear.com/product/testcomplete, Jan,
2015.
[17] http://www.testplant.com, Jan, 2015.
[18] http://www.froglogic.com, Jan, 2015.
[19] O. S. Pietro, H. Jun, Y. Xin, "Time complexity
of evolutionary algorithms for combinatorial
optimization: A decade of results",
International Journal of Automation and
Computing, vol. 4, pp. 218-293, 2007.
[20] https://www.tizen.org, Jan, 2015.
E-ISSN: 2224-2872 579 Volume 14, 2015

Hidden Object Detection for Computer Vision Based Test Automation System

More Related Content

Viewers also liked (8)

Similar to Hidden Object Detection for Computer Vision Based Test Automation System (20)

More from Syeful Islam (15)

Recently uploaded (20)

Hidden Object Detection for Computer Vision Based Test Automation System