DETECTION OF BUILDING CRACKS FROM IMAGES IN CLOUD-FOG COMPUTING

Papers 2023

TITLE
DETECTION OF BUILDING CRACKS FROM IMAGES IN CLOUD-FOG COMPUTING

AUTHOR(S)
Dušan Marković, Dejan Vujičić, Dijana Stojić, Uroš Pešović, Siniša Ranđić

ABSTRACT
Inspection of infrastructures, such as buildings, is significant to detect defects that can cause more damage. Finding defects, such as cracks on the building surface, timely represents information that helps to maintain stability, safety, and duration of the building. Certain parts of the building surface may be difficult to access manually. So, an automated system could be used with unmanned aerial vehicles (UAV) and computer vision techniques coupled with Convolutional Neural Networks (CNNs). Our work aims to present the process of training a neural network model for building crack detection on the Cloud with satisfactory accuracy. And then deploy that model at the Fog level for the new image classification. The subject of research is the distribution of processing tasks from a distance server or Cloud to the Fog node closer to the source to obtain the results of image processing without high value for delay, reduce data transmission to the Cloud platform, and thus reduce the network load and energy consumption on the Cloud.

DOI
www.doi.org/10.70456/ZULJ4244

PAGES
167-172

DOWNLOAD
https://unitechsp.tugab.bg/images/2023/4-KS/s5_p172_v1.pdf

How to cite this article:
Dušan Marković, Dejan Vujičić, Dijana Stojić, Uroš Pešović, Siniša Ranđić, DETECTION OF BUILDING CRACKS FROM IMAGES IN CLOUD-FOG COMPUTING, UNITECH – SELECTED PAPERS - 2024, 167-172

USING IMAGE PROCESSING TECHNIQUES FOR PART CLASSIFICATION

Papers 2023

TITLE
USING IMAGE PROCESSING TECHNIQUES FOR PART CLASSIFICATION

AUTHOR(S)
Aydın GULLU, M. Ozan AKI

ABSTRACT
In industrial automation, it is used with various sensors for detection. Digital sensors are widely used because they are economical and easy to process. There are digital sensors that detect various materials for part detection. There is a FESTO part sorting station at Trakya University İpsala Vocational School. At the end of the production line, the color and structure of the part is detected and classified by industrial sensors. The part is detected by two optical sensors and the color of the part is determined according to the light reflected from the part. After color detection, it is checked whether the piece is metal or not by means of an inductive sensor. According to the information coming from these three sensors, the parts are classified as black, red and metallic. Instead of this
sensor combination, a camera was used in this study. The color of the part is detected by image processing. In addition, thanks to the camera, part thickness and diameter were also detected. Classification and quality control of the parts were provided with the camera. In the study, the camera works connected to the computer. The classification station is controlled by PLC. Image processing data was transferred to the PLC with a microcontroller card.

DOI
www.doi.org/10.70456/VCJB4824

PAGES
161-166

DOWNLOAD
https://unitechsp.tugab.bg/images/2023/4-KS/s5_p168_v2.pdf

How to cite this article:
Aydın GULLU, M. Ozan AKI, USING IMAGE PROCESSING TECHNIQUES FOR PART CLASSIFICATION, UNITECH – SELECTED PAPERS - 2024, 161-166

QUALITY CONTROL OF PARTS IN INDUSTRIAL PRODUCTION WITH IMAGE PROCESSING METHODS

Papers 2023

TITLE
QUALITY CONTROL OF PARTS IN INDUSTRIAL PRODUCTION WITH IMAGE PROCESSING METHODS

AUTHOR(S)
M. Ozan AKI, Aydın GULLU

ABSTRACT
Mass production is done with industrial machines. During manufacturing, the parts undergo various processes to give them their final shape. In machining, the processing of the part is inspected at the final stage. In this study, a visual inspection system has been designed to determine the quality of the part at the final stage of processing. The semi-finished cylindrical parts are processed on the rotating round magazine in the current production system.
The six-step round magazine has four processing stops which are take-in, centering, drilling, and take-out the parts consecutively. The developed visual inspection system in this study is designed for the last stop of this round magazine. The part images are captured by the camera and the color and diameter of the part and areas of holes on the part are determined. The developed inspection system makes pass or failed decisions for each part by
comparing measurements with the specifications of the part. Measurements of all parts are saved to the process database for further analysis. As a result of this study, the developed inspection system is considered suitable and integratable for round magazine process quality control.

DOI
www.doi.org/10.70456/TCCG5054

PAGES
155-160

DOWNLOAD
https://unitechsp.tugab.bg/images/2023/4-KS/s5_p167_v1.pdf

How to cite this article:
M. Ozan AKI, Aydın GULLU, QUALITY CONTROL OF PARTS IN INDUSTRIAL PRODUCTION WITH IMAGE PROCESSING METHODS, UNITECH – SELECTED PAPERS - 2024, 155-160

APPLICATION OF NEURAL NETWORKS IN ANDROID APPLICATIONS FOR OBJECT RECOGNITION IN REAL TIME

Papers 2023

TITLE
APPLICATION OF NEURAL NETWORKS IN ANDROID APPLICATIONS FOR OBJECT RECOGNITION IN REAL TIME

AUTHOR(S)
Zeljko Jovanovic, Filip Petrovic, Mihailo Knezevic

ABSTRACT
This paper describes how neural networks can be used in Android applications. Specifically, an educational application for language learning based on a neural network model was developed. Several neural network models were trained for object detection as part of the practical part. With the help of these models, an
application that detects objects in real-time and translates them into the desired language was created. Several topics were explored in this paper, such as neural networks, artificial intelligence, android platform, TensorFlow and TensorFlow Lite libraries and how they work, and the concept of detection, i.e., object
recognition. In addition to the theoretical part, which is necessary to understand how neural networks work and the Android platform, each step of the practical demonstration is described in detail. This includes preparing the working environment, training the data set, training the neural network, and developing the Android application.

DOI
www.doi.org/10.70456/IGXF8861

PAGES
149-154

DOWNLOAD
https://unitechsp.tugab.bg/images/2023/4-KS/s5_p78_v4.pdf

How to cite this article:
Zeljko Jovanovic, Filip Petrovic, Mihailo Knezevic, APPLICATION OF NEURAL NETWORKS IN ANDROID APPLICATIONS FOR OBJECT RECOGNITION IN REAL TIME, UNITECH – SELECTED PAPERS - 2024, 149-154

DEVELOPMENT OF AN IP CAMERA SYSTEM USING MACHINE LEARNING FOR THE PURPOSE OF PRIVACY PROTECTION

Papers 2023

TITLE
DEVELOPMENT OF AN IP CAMERA SYSTEM USING MACHINE LEARNING FOR THE PURPOSE OF PRIVACY PROTECTION

AUTHOR(S)
Oliver Popović, Nikola Nikolić, Vladica Ubavić, Marina Jovanović-Milenković, Marko Asanović, Ivana Buzdovan

ABSTRACT
The rapid evolution of technology and the Internet have profoundly influenced the advancement of video surveillance systems. In contemporary video surveillance systems, camera-recorded content has the potential to be publicly accessible via the Internet. However, the widespread deployment of such cameras has raised concerns about privacy violations, as these devices capture extensive footage of individuals and activities. To address these
concerns and safeguard privacy, various technologies and software solutions have been deployed. This paper elaborates the architectural design and development of a system that use machine learning techniques to obfuscate individuals' faces within recorded materials. By employing machine learning, the proposed system ensures that the identity of individuals remains protected, both in publicly available content and in cases of unauthorized
access, granting exclusive access to the original, unobscured content only to authorized personnel.

DOI
www.doi.org/10.70456/TKVC5411

PAGES
144-148

DOWNLOAD
https://unitechsp.tugab.bg/images/2023/4-KS/s5_p38_v3.pdf

How to cite this article:
Oliver Popović, Nikola Nikolić, Vladica Ubavić, Marina Jovanović-Milenković, Marko Asanović, Ivana Buzdovan, DEVELOPMENT OF AN IP CAMERA SYSTEM USING MACHINE LEARNING FOR THE PURPOSE OF PRIVACY PROTECTION, UNITECH – SELECTED PAPERS - 2024, 144-148