• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.2
• /
• pp.77-83
• /
• 2022

In this paper, the two-stage object detection approach is proposed to implement a deep learning-based label inspection system on edge computing environments. Since the label printed on the products during the production process contains important information related to the product, it is significantly to check the label information is correct. The proposed system uses the lightweight deep learning model that able to employ in the low-performance edge computing devices, and the two-stage object detection approach is applied to compensate for the low accuracy relatively. The proposed Two-Stage object detection approach consists of two object detection networks, Label Area Detection Network and Character Detection Network. Label Area Detection Network finds the label area in the product image, and Character Detection Network detects the words in the label area. Using this approach, we can detect characters precise even with a lightweight deep learning models. The SF-YOLO model applied in the proposed system is the YOLO-based lightweight object detection network designed for edge computing devices. This model showed up to 2 times faster processing time and a considerable improvement in accuracy, compared to other YOLO-based lightweight models such as YOLOv3-tiny and YOLOv4-tiny. Also since the amount of computation is low, it can be easily applied in edge computing environments.

• ETRI Journal
• /
• v.25 no.5
• /
• pp.401-406
• /
• 2003

A biometric system determines the identity of a person by measuring physical features that can distinguish that person from others. Since biometric features have many variations and can be easily corrupted by noises and deformations, it is necessary to apply machine learning techniques to treat the data. When applying the conventional machine learning methods in designing a specific biometric system, however, one first runs into the difficulty of collecting sufficient data for each person to be registered to the system. In addition, there can be an almost infinite number of variations of non-registered data. Therefore, it is difficult to analyze and predict the distributional properties of real data that are essential for the system to deal with in practical applications. These difficulties require a new framework of identification and verification that is appropriate and efficient for the specific situations of biometric systems. As a preliminary solution, this paper proposes a simple but theoretically well-defined method based on a statistical test theory. Our computational experiments on real-world data show that the proposed method has potential for coping with the actual difficulties in biometrics.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.14 no.6
• /
• pp.295-311
• /
• 2019

The smart product market is growing year by year and is being used in many areas. There are various ways of interacting with smart products and users by inputting voice recognition, touch and finger movements. It is most important to detect an accurate hand region as a whole step to recognize hand movement. In this paper, we propose a method to detect accurate hand region in real time in various environments. A conventional method of detecting a hand region includes a method using depth information of a multi-sensor camera, a method of detecting a hand through machine learning, and a method of detecting a hand region using a color model. Among these methods, a method using a multi-sensor camera or a method using a machine learning requires a large amount of calculation and a high-performance PC is essential. Many computations are not suitable for embedded systems, and high-end PCs increase or decrease the price of smart products. The algorithm proposed in this paper detects the hand region using the color model, corrects the problems of the existing hand detection algorithm, and detects the accurate hand region based on various experimental environments.

• Journal of KIISE:Software and Applications
• /
• v.32 no.9
• /
• pp.920-926
• /
• 2005

Nearest neighbor algorithms classify an unseen input instance by selecting similar cases and use the discovered membership to make predictions about the unknown features of the input instance. The usefulness of the nearest neighbor algorithms have been demonstrated sufficiently in many real-world domains. In nearest neighbor algorithms, it is an important issue to assign proper weights to the attributes. Therefore, in this paper, we propose a new method which can automatically assigns to each attribute a weight of its importance with respect to the target attribute. The method has been implemented as a computer program and its effectiveness has been tested on a number of machine learning databases publicly available.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.31 no.3
• /
• pp.365-371
• /
• 2021

With the rapid growth in Internet users, web applications are becoming the main target of hackers. Most previous WAFs (Web Application Firewalls) target every single HTTP request packet rather than the overall behavior of the attacker, and are known to be difficult to detect new types of attacks. In this paper, we propose a web attack detection system based on user behavior using machine learning to detect attacks of unknown patterns. In order to define user behavior, we focus on features excluding areas where an attacker can camouflage as a normal user. The experimental results shows that by using the path and query information to define users' behaviors, best results for an accuracy of 99% with Decision forest.

• Advances in nano research
• /
• v.12 no.4
• /
• pp.387-403
• /
• 2022

Silica nanoparticles, which have a broad range of sizes and specific surface features, have been used in many industrial applications. This study was conducted to synthesize monodispersed silica nanoparticles directly from tetraethyl orthosilicate (TEOS) with an alkaline catalyst (NH₃) based on the sol-gel process and the Stöber method. A central composite design (CCD) is used to build a second-order (quadratic) model for the response variables without requiring a complete three-level factorial experiment. The process was then optimized to achieve the minimum particle size with the lowest concentration of TEOS. Dynamic light scattering and scanning electron microscopy were used to analyze the size, dispersity, and morphology of the synthesized nanoparticles. After optimization, a confirmation test was carried out to evaluate the confidence level of the software prediction. The results revealed that the predicted optimization is consistent with experimental procedures, and the model is significant at the 95% confidence level.

• Steel and Composite Structures
• /
• v.43 no.4
• /
• pp.501-509
• /
• 2022

Cracks are common defects in concrete structures. Thus far, crack inspection has been manually performed using the contact inspection method. This manpower-dependent method inevitably increases the cost and work hours. Various non-contact studies have been conducted to overcome such difficulties. However, previous studies have focused on developing a methodology for non-contact inspection or local quantitative detection of crack width or length on concrete surfaces. However, crack depth can affect the safety of concrete structures. In particular, although macrocrack depth is structurally fatal, it is difficult to find it with the existing method. Therefore, an experimental investigation based on non-contact infrared thermography and multivariate machine learning was performed in this study to estimate the hidden macrocrack depth. To consider practical applications for inspection, an experiment was conducted that considered the simulated piloting of an unmanned aerial vehicle equipped with infrared thermography equipment. The crack depths (10-60 mm) were comparatively evaluated using linear regression, gradient boosting, and random forest (AI regression methods).

• Structural Engineering and Mechanics
• /
• v.84 no.1
• /
• pp.101-111
• /
• 2022

It is widely known that axially loaded fiber-reinforced polymer (FRP) confined concrete presents significant and enhanced mechanical properties with reference to the unconfined concrete. Therefore, to predict the mechanical behavior of FRP-confined concrete two quantities-peak strength and ultimate strain are required. Despite the significant advances, the determination of the ultimate strain of FRP-confined concrete is one of the most challenging problems to be resolved. This is often attributed to our persistence in desiring the conventional methods as the sole technique to examine this phenomenon and the complex nature of the ultimate strain of FRP-confined concrete. To bridge the research gap, this study adopted two machine learning (ML) techniques-artificial neural network (ANN) and Gaussian process regression (GPR)-to analyze observations obtained from 627 datasets of FRP-confined concrete circular and non-circular sections under axial loading test. Besides, the techniques are also used to predict the ultimate strain of FRP-confined concrete. Seven parameters namely width/diameter of the specimens, corner radius ratio, the strength of concrete, FRP elastic modulus, FRP thickness, FRP tensile rupture strain, and the axial strain of unconfined concrete-are the input parameters used to predict the ultimate strain of FRP-confined concrete. The results of the current study highlight the merit of using AI techniques in structural engineering applications given their extraordinary ability to comprehend multidimensional phenomena of FRP-confined concrete structures with ease, low computational cost, and high performance over the existing empirical models.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.7
• /
• pp.1916-1934
• /
• 2023

The phishing attack is a malicious emerging threat on the internet where the hackers try to access the user credentials such as login information or Internet banking details through pirated websites. Using that information, they get into the original website and try to modify or steal the information. The problem with traditional defense systems like firewalls is that they can only stop certain types of attacks because they rely on a fixed set of principles to do so. As a result, the model needs a client-side defense mechanism that can learn potential attack vectors to detect and prevent not only the known but also unknown types of assault. Feature selection plays a key role in machine learning by selecting only the required features by eliminating the irrelevant ones from the real-time dataset. The proposed model uses Hyperparameter Optimized Artificial Neural Networks (H-ANN) combined with a Hybrid Firefly and Grey Wolf Optimization algorithm (H-FFGWO) to detect and block phishing websites in Internet of Things(IoT) Applications. In this paper, the H-FFGWO is used for the feature selection from phishing datasets ISCX-URL, Open Phish, UCI machine-learning repository, Mendeley website dataset and Phish tank. The results showed that the proposed model had an accuracy of 98.07%, a recall of 98.04%, a precision of 98.43%, and an F1-Score of 98.24%.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.4
• /
• pp.751-756
• /
• 2024

The accuracy of the model and the explanation of the results are important factors that should be considered simultaneously Recently, applications of explainable artificial intelligence are increasing, and it is especially widely applied in the financial field where interpretation of results is important. In this paper, we compare the performance of open API credit evaluation data using various machine learning techniques. In addition, existing financial logic is verified through explainable artificial intelligence technologies, SHAP and LIME. Accordingly, it is expected to demonstrate the applicability of machine learning in the financial market.