Demand for OIS (Optical Image Stabilization) actuator modules, developed for shake correction technologies in industries such as smartphones, drones, IoT, and AR/VR, is increasing. To enable real-time and precise inspection of these modules, an AI algorithm that maximizes defect detection accuracy is required. This study proposes an unsupervised learning-based algorithm that is robust to noise and capable of real-time processing for accurate defect classification of OIS actuators in a smart factory environment. The proposed algorithm performs noise-reduction preprocessing, considering the sensitivity of small components and lighting imbalances, and defines three dynamic Regions of Interest (ROIs) to address positional deviations. A customized AutoEncoder (AE) is trained for each ROI, and defect classification is conducted based on reconstruction errors, followed by a final comprehensive decision. Experimental results show that the algorithm achieves an accuracy of 0.9944 and an F1 score of 0.9971 using only a camera without the need for expensive sensors. Furthermore, it demonstrates an average processing time of 2.79 ms per module, ensuring real-time capability. This study contributes to precise quality inspection in smart factories by proposing a robust and scalable unsupervised inspection algorithm.

With the growth of e-commerce platforms, the number of fraud reviews has rapidly increased, making fraud detection for platform data critically important. To detect fraud reviews in platform data, recent approaches have leveraged graph neural networks that model users as nodes in a heterogeneous multi-relational graph structure. This approach represents platform users as nodes in a graph, and the relationships among users who share commonalities in their reviews or products as edges, in order to identify fraudulent reviewers. However, existing graph-based fraud detection methods may suffer from the unstable training of the classifier networks that process embedding vectors. In this paper, we identify that this issue arises from the excessive deviation and scale expansion caused by some of the aggregated values from adjacent nodes in conventional node embeddings, and we propose a scale regularization method to mitigate this. To verify the effectiveness of the proposed method, we conduct validation on the Amazon-Fraud dataset, which is a multi-relational graph dataset constructed from review data of Amazon E-Commerce. The experimental results show that the proposed scale regularization achieves superior performance compared to previous verified graph fraud detection models.

In the manufacturing supply chain of smart devices, defect inspection for core components, such as lens actuator that constitute the final product, is essential for ensuring productivity and reliability. Especially, the ball-based lens actuator—which controls the movement of the lens through balls mounted in the camera module—is one of the essential devices that require inspection of structural features such as the position, number, and arrangement state of the balls. In the industrial component inspection domain, general-purpose anomaly detection model based on deep learning methods have primarily been used as a means of detecting visual defects; however, they are not suitable for inspecting anomalies in structural features, such as those found in lens actuator. For this reason, this study proposes an anomaly detection process based on structural features. In this study, the proposed method is designed to reduce false positives in defect inspection by explicitly detecting structural anomalies through pattern recognition and computer vision techniques. In this paper, the proposed method was applied to various types of experiments using a dataset of images collected from an actual actuator assembly process. The proposed method showed that it demonstrated superior performance, as well as robustness in invariant-feature recognition compared to general-purpose anomaly detection models. (13.75% higher precision than the conventional inspection system, and 1.48% to 35.84% higher recall than the conventional general-purpose anomaly detection models).

In the modern digital age, users are exposed to a vast amount of content and information, and the importance of recommendation systems is increasing accordingly. Traditional recommendation systems mainly use matrix factorization and collaborative filtering methods, but problems with scalability due to an increase in the amount of data and slow learning and inference speeds occur due to an increase in the amount of computation. To overcome these problems, this study focused on optimizing LightGCN, the basic structure of the graph-convolution-network-based recommendation system. To improve this, techniques and structures were proposed. We propose an embedding enhancement method to strengthen the robustness of embedding and a non-combination structure to overcome LightGCN’s weight sum structure through this method. To verify the proposed method, we have demonstrated its effectiveness through experiments using the SELFRec library on various datasets, such as Yelp2018, MovieLens-1M, FilmTrust, and Douban-book. Mainly, significant performance improvements were observed in key indicators, such as Precision, Recall, NDCG, and Hit Ratio in Yelp2018 and Douban-book datasets. These results suggest that the proposed methods effectively improved the recommendation performance and learning efficiency of the LightGCN model, and the improvement of LightGCN, which is most widely used as a backbone network, makes an important contribution to the entire field of GCN-based recommendation systems. Therefore, in this study, we improved the learning method of the existing LightGCN and changed the weight sum structure to surpass the existing accuracy.

This research aims to propose a practical framework designed for the automatic analysis of a product's comprehensive functionality and security vulnerabilities, generating applicable guidelines based on real-world software. The existing analysis of software security vulnerabilities often focuses on specific features or modules. This partial and arbitrary analysis of the security vulnerabilities makes it challenging to comprehend the overall security vulnerabilities of the software. The key novelty lies in overcoming the constraints of partial approaches. The proposed framework utilizes data from various sources to create a comprehensive functionality profile, facilitating the derivation of real-world security guidelines. Security guidelines are dynamically generated by associating functional security vulnerabilities with the latest Common Vulnerabilities and Exposure (CVE) and Common Vulnerability Scoring System (CVSS) scores, resulting in automated guidelines tailored to each product. These guidelines are not only practical but also applicable in real-world software, allowing for prioritized security responses. The proposed framework is applied to virtual private network (VPN) software, wherein a validated Level 2 data flow diagram is generated using the Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, and Elevation of privilege (STRIDE) technique with references to various papers and examples from related software. The analysis resulted in the identification of a total of 121 vulnerabilities. The successful implementation and validation demonstrate the framework's efficacy in generating customized guidelines for entire systems, subsystems, and selected modules.

When firefighters search inside a building that is at risk of collapse due to abandonment or disasters such as fire, they use old architectural drawings or a simple monitoring method involving a video device attached to a robot. However, using these methods, the disaster situation inside a building at risk of collapse is difficult to detect and identify. Therefore, we investigate the generation of digital maps for a disaster site to accurately analyze internal situations. In this study, a robot combined with a low-cost camera and twodimensional light detection and ranging (2D-lidar) traverses across a floor to estimate the location of obstacles while drawing an internal map of the building. We propose an algorithm that detects the floor and then determines the possibility of entry, tracks collapses, and detects obstacles by analyzing patterns on the floor. The robot???s location is estimated, and a digital map is created based on Hector simultaneous localization and mapping (SLAM). Subsequently, the positions of obstacles are estimated based on the range values detected by 2D-lidar, and the position of the obstacles are identified on the map using the map update method in semantic SLAM. All equipment are implemented using low-specification devices, and the experiments are conducted using a low-cost robot that affords near-real-time performance. The experiments are conducted in various actual internal environments of buildings. In terms of obstacle detection performance, almost all obstacles are detected, and their positions identified on the map with a high accuracy of 89%.

Existing firefighting robots are focused on simple storage or fire suppression outside buildings rather than detection or recognition. Utilizing a large number of robots using expensive equipment is challenging. This study aims to increase the efficiency of search and rescue operations and the safety of firefighters by detecting and identifying the disaster site by recognizing collapsed areas, obstacles, and rescuers on-site. A fusion algorithm combining a camera and three-dimension light detection and ranging (3D LiDAR) is proposed to detect and localize the interiors of disaster sites. The algorithm detects obstacles by analyzing floor segmentation and edge patterns using a mask regional convolutional neural network (mask R-CNN) features model based on the visual data collected from a parallelly connected camera and 3D LiDAR. People as objects are detected using you only look once version 4 (YOLOv4) in the image data to localize persons requiring rescue. The point cloud data based on 3D LiDAR cluster the objects using the density-based spatial clustering of applications with noise (DBSCAN) clustering algorithm and estimate the distance to the actual object using the center point of the clustering result. The proposed artificial intelligence (AI) algorithm was verified based on individual sensors using a sensor-mounted robot in an actual building to detect floor surfaces, atypical obstacles, and persons requiring rescue. Accordingly, the fused AI algorithm was comparatively verified.

When firefighters are engaged in search and rescue missions inside a building at a risk of collapse, they have difficulty in field command and rescue because they can only simply monitor the situation inside the building utilizing old building drawings or robots.To propose an efficient solution for fast search and rescue work of firefighters, this study investigates the generation of up-to-date digital maps for disaster sites by tracking the collapse situation, and identifying the information of obstacles which are risk factors, using an artificial intelligence algorithm based on low-cost robots.Our research separates the floor by using the mask regional convolutional neural network (R-CNN) algorithm, and determines whether the passage is collapsed or not.Then, in the case of a passage that can be searched, the floor pattern of the obstacles that exist on the floor that has not collapsed is analyzed, and obstacles are searched utilizing an image processing algorithm.Here, we can detect various unknown as well as known obstacles.Furthermore, the locations of obstacles can be estimated using the pixel values up to the bounding box of an existing detected obstacle.We conduct experiments using the public datasets collected by Carnegie Mellon university (CMU) and data collected by manipulating a low-cost robot equipped with a smartphone while roaming five buildings in a campus.The collected data have various floor patterns for objectivity and obstacles that are different from one another.Based on these data, the algorithm for detecting unknown obstacles of a verified study and estimating their sizes had an accuracy of 93%, and the algorithm for estimating the distance to obstacles had an error rate of 0.133.Through this process, we tracked collapsed passages and composed up-to-date digital maps for disaster sites that include the information of obstacles that interfere with the search and rescue work.

Wildlife ecologists and biologists recapture deer to collect tracking data from deer collars or wait for a drop-off of a deer collar construction that is automatically detached and disconnected. The research teams need to manage a base camp with medical trailers, helicopters, and airplanes to capture deer or wait for several months until the deer collar drops off of the deer's neck. We propose an intelligent robust base-station research with a low-cost and time saving method to obtain recording sensor data from their collars to a listener node, and readings are obtained without opening the weatherproof deer collar. We successfully designed the and implemented a robust base station system for automatically collecting data of the collars and listener motes in harsh wilderness environments. Intelligent solutions were also analyzed for improved data collections and pattern predictions with drone-based detection and tracking algorithms.

Due to the significant increase in the use of autonomous car technology, it is essential to integrate this technology with high-precision digital map data containing more precise and accurate roadway information, as compared to existing conventional map resources, to ensure the safety of self-driving operations. While existing map technologies may assist vehicles in identifying their locations via Global Positioning System, it is however difficult to update the environmental changes of roadways in these maps. Roadway vision algorithms can be useful for building autonomous vehicles that can avoid accidents and detect real-time location changes. We incorporate a hybrid architectural design that combines unsupervised classification of vision data with supervised joint fusion classification to achieve a better noise-resistant algorithm. We identify, via a deep learning approach, an intelligent hybrid fusion algorithm for fusing multimodal vision feature data for roadway classifications and characterize its improvement in accuracy over unsupervised identifications using image processing and supervised vision classifiers. We analyzed over 93,000 vision frame data collected from a test vehicle in real roadways. The performance indicators of the proposed hybrid fusion algorithm are successfully evaluated for the generation of roadway digital maps for autonomous vehicles, with a recall of 0.94, precision of 0.96, and accuracy of 0.92.

According to the statistics examined by the National Safety Council, Injury Facts in 2017, a significant number of preventable injury-related deaths occurred in home and indoor public areas, and the rate of preventable deaths occurring indoors has increased by 156% since 1999. Indoor smart Internet of Things devices such as security cameras, intelligent speakers, smartphones, and air cleaners are being utilized to seek help during dangerous or emergency situations in indoor settings. Theses Internet of Things devices can also assist single-person household rescues during emergency situations where no one is present for assistance. We propose a real-time algorithm to detect unusual user events which may help in reducing the rate of people dying alone and remaining undiscovered for a long period of time. We designed and developed unusual user behavior patterns related to vision, audio, dust, and activity via Internet of Things sensors and fused these patterns to improve the performance accuracy. We evaluated the proposed individual pattern algorithms and the fusion method through the data collected in indoor smart environments.

The net profit of investors can rapidly increase if they correctly decide to take one of these three actions: buying, selling, or holding the stocks. The right action is related to massive stock market measurements. Therefore, defining the right action requires specific knowledge from investors. The economy scientists, following their research, have suggested several strategies and indicating factors that serve to find the best option for trading in a stock market. However, several investors' capital decreased when they tried to trade the basis of the recommendation of these strategies. That means the stock market needs more satisfactory research, which can give more guarantee of success for investors. To address this challenge, we tried to apply one of the machine learning algorithms, which is called deep reinforcement learning (DRL) on the stock market. As a result, we developed an application that observes historical price movements and takes action on real-time prices. We tested our proposal algorithm with three-Bitcoin (BTC), Litecoin (LTC), and Ethereum (ETH)-crypto coins' historical data. The experiment on Bitcoin via DRL application shows that the investor got 14.4% net profits within one month. Similarly, tests on Litecoin and Ethereum also finished with 74% and 41% profit, respectively.

A large number of studies that use artificial intelligence (AI) methodologies to analyze medical imaging and support computer-aided diagnosis have been conducted in the biomedical engineering domain. Owing to the advances in dental diagnostic X-ray systems such as panoramic radiographs, periapical radiographs, and dental computed tomography (CT), especially, dual-energy cone beam CT (CBCT), dental image analysis now presents more opportunities to discover new results and findings. Recent researches on dental image analysis have been increasingly incorporating analytics that utilize AI methodologies that can be divided into conventional machine learning and deep learning approaches. This review first covers the theory on dual-energy CBCT and its applications in dentistry. Then, analytical methods for dental image analysis using conventional machine learning and deep learning methods are described. We conclude by discussing the issues and suggesting directions for research in future.

The grid map, representing area information with the number of cells, is a widely used mapping scheme for mobile robots and simultaneous localization and mapping (SLAM) processes. However, the tremendous amount of cells in a grid map for a detailed map representation results in overheads for memory space and computing paths in mobile robots. Therefore, to overcome the overhead of the grid map, this study proposes a new low overhead mapping scheme which the authors call as the Rmap that represents an area with variable sizes of rectangles instead of the number of cells in the grid map. This mapping scheme also provides an exploration path for obtaining new information for the unknown area. This study evaluated the performance of the Rmap in real environments as well as in simulation environments. The experiment results show that the Rmap can reduce the overhead of a grid map. In one of our experimental environments, the Rmap represented an area with 85% less memory than the grid map. The Rmap also showed better coverage performance compared with other previous algorithms.

Interest in self-driving vehicle research has been rapidly increasing, and related research has been continuously conducted. In such a fast-paced self-driving vehicle research area, the development of advanced technology for better convenience safety, and efficiency in road and transportation systems is expected. Here, we investigate research in self-driving vehicles and analyze the main technologies of driverless car software, including: technical aspects of autonomous vehicles, traffic infrastructure and its communications, research techniques with vision recognition, deep leaning algorithms, localization methods, existing problems, and future development directions. First, we introduce intelligent self-driving car and road infrastructure algorithms such as machine learning, image processing methods, and localizations. Second, we examine the intelligent technologies used in self-driving car projects, autonomous vehicles equipped with multiple sensors, and interactions with transport infrastructure. Finally, we highlight the future direction and challenges of self-driving vehicle transportation systems.

According to the Statistics Korea in 2017, the 10 leading causes of death contain a cardiac disorder disease, self-injury. In terms of these diseases, urgent assistance is highly required when people do not move for certain period of time. We propose an unusual event detection algorithm to identify abnormal user behaviors using dust, vision and activity sensors in their houses. Vision sensors can detect personalized activity behaviors within the CCTV range in the house in their lives. The pattern algorithm using the dust sensors classifies user movements or dust-generated daily behaviors in indoor areas. The accelerometer sensor in the smartphone is suitable to identify activity behaviors of the mobile users. We evaluated the proposed pattern algorithms and the fusion method in the scenarios. ▸Keyword: Dust, Vision, Activity, Pattern recognition, Artificial intelligence