• Journal of Intelligence and Information Systems
• /
• v.28 no.1
• /
• pp.89-106
• /
• 2022

Recently, the amount of video data collected from smartphones, CCTVs, black boxes, and high-definition cameras has increased rapidly. According to the increasing video data, the requirements for analysis and utilization are increasing. Due to the lack of skilled manpower to analyze videos in many industries, machine learning and artificial intelligence are actively used to assist manpower. In this situation, the demand for various computer vision technologies such as object detection and tracking, action detection, emotion detection, and Re-ID also increased rapidly. However, the object detection and tracking technology has many difficulties that degrade performance, such as re-appearance after the object's departure from the video recording location, and occlusion. Accordingly, action and emotion detection models based on object detection and tracking models also have difficulties in extracting data for each object. In addition, deep learning architectures consist of various models suffer from performance degradation due to bottlenects and lack of optimization. In this study, we propose an video analysis system consists of YOLOv5 based DeepSORT object tracking model, SlowFast based action recognition model, Torchreid based Re-ID model, and AWS Rekognition which is emotion recognition service. Proposed model uses single-linkage hierarchical clustering based Re-ID and some processing method which maximize hardware throughput. It has higher accuracy than the performance of the re-identification model using simple metrics, near real-time processing performance, and prevents tracking failure due to object departure and re-emergence, occlusion, etc. By continuously linking the action and facial emotion detection results of each object to the same object, it is possible to efficiently analyze videos. The re-identification model extracts a feature vector from the bounding box of object image detected by the object tracking model for each frame, and applies the single-linkage hierarchical clustering from the past frame using the extracted feature vectors to identify the same object that failed to track. Through the above process, it is possible to re-track the same object that has failed to tracking in the case of re-appearance or occlusion after leaving the video location. As a result, action and facial emotion detection results of the newly recognized object due to the tracking fails can be linked to those of the object that appeared in the past. On the other hand, as a way to improve processing performance, we introduce Bounding Box Queue by Object and Feature Queue method that can reduce RAM memory requirements while maximizing GPU memory throughput. Also we introduce the IoF(Intersection over Face) algorithm that allows facial emotion recognized through AWS Rekognition to be linked with object tracking information. The academic significance of this study is that the two-stage re-identification model can have real-time performance even in a high-cost environment that performs action and facial emotion detection according to processing techniques without reducing the accuracy by using simple metrics to achieve real-time performance. The practical implication of this study is that in various industrial fields that require action and facial emotion detection but have many difficulties due to the fails in object tracking can analyze videos effectively through proposed model. Proposed model which has high accuracy of retrace and processing performance can be used in various fields such as intelligent monitoring, observation services and behavioral or psychological analysis services where the integration of tracking information and extracted metadata creates greate industrial and business value. In the future, in order to measure the object tracking performance more precisely, there is a need to conduct an experiment using the MOT Challenge dataset, which is data used by many international conferences. We will investigate the problem that the IoF algorithm cannot solve to develop an additional complementary algorithm. In addition, we plan to conduct additional research to apply this model to various fields' dataset related to intelligent video analysis.

• The Journal of Engineering Geology
• /
• v.23 no.4
• /
• pp.353-361
• /
• 2013

To analyze the distribution of earthquake-induced strain data in rock masses, statistical analysis was performed on four-directional strain data obtained from a ground movement monitoring system installed in Korea. Strain data related to the 2011 Tohoku-oki earthquake and two aftershocks of >M7.0 in 2011 were used in x-MR control chart analysis, a type of univariate statistical analysis that can detect an abnormal distribution. The analysis revealed different dispersion times for each measurement orientation. In a more comprehensive analysis, the strain data were re-evaluated using multivariate statistical analysis (MSA) considering correlations among the various data from the different measurement orientations. $T_2$ and Q-statistics, based on principal component analysis, were used to analyze the time-series strain data in real-time. The procedures were performed with 99.9%, 99.0%, and 95.0% control limits. It is possible to use the MSA data to successfully detect an abnormal distribution caused by earthquakes because the dispersion time using the 99.9% control limit is concurrent with or earlier than that from the x-MR analysis. In addition, the dispersion using the 99.0% and 95.0% control limits detected an abnormal distribution in advance. This finding indicates the potential use of MSA for recognizing abnormal distributions of strain data.

• Journal of Internet Computing and Services
• /
• v.10 no.1
• /
• pp.19-31
• /
• 2009

Network management involves configuring and operating various network elements in a suitable manner. Generally, a network management system can perform basic functionalities such as configuration management, performance management, and fault management. Due to the open structure of the Internet, the volume of network traffic and the network equipment used have increased in size and complexity. Therefore, it is expensive and time consuming to develop a network management program for heterogeneous network equipment in an SNMP.based network. In order to facilitate the management of network environments and the control of heterogeneous devices in an efficient manner, we propose an Active Network Management System (ANMS) comprising an automatic generator that uses triggers to generate a network management program. The concept of triggers can be represented through event condition action rules performed in response to a change in the status of a network environment. The proposed ANMS comprises basic components for real time network management and also includes an automatic generator (AG). When the ANMS is monitoring network elements that are newly added or changed, a trigger rule is activated and these components are then able to collaborate and automatically generate a new network management program by using the information provided along with the SNMP libraries. Our method is useful for expanding the network structure and replacing network equipment. Through experiments, we have proved that our ANMS is useful when new network objects are added or changed in the network environment to expand the network structure. Further, we have verified that our ANMS system reduces the time and cost required to develop a network management program as compared to the manual method used in existing network management systems.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.14 no.1
• /
• pp.259-264
• /
• 2014

IEEE 802.15.4 has been emerging as the popular choice for various monitoring and control applications. In this paper, a fault management system for DC power-lines has been designed using IEEE 802.15.4, in order to monitor DC power-lines in real time, and to rapidly detect faults and shut off the line where such faults occur. Numbers were allocated for each node and unslotted CSMA-CA method of IEEE 802.15.4 was used, the performance of which was analyzed by a simulation. For such purpose, a total of 60 bits of the control data consisting of 16 bits of the current, 16 bits of the amplitude, 28 bits of the terminal state data were sent out, and the packet transfer rate and the transmission delay time of the fault management system for DC power-lines were measured and analyzed. When the traffic load was 330 packets per second or lower, the average delay time was shown to be shorter than 0.02 seconds, and when the traffic load was 260 packets per second or lower, the packet transfer rate was shown to be 99.99% or higher. Therefore, it was confirmed that the stringent condition of US Department of Energy (DOE) could be satisfied if the traffic load was 260 packets per second or lower, The results of this study can be utilized as basic data for the establishment of the fault management system for DC power-lines using IEEE 802.15.4.

• Journal of Environmental Health Sciences
• /
• v.44 no.1
• /
• pp.98-105
• /
• 2018

Objectives: In Mongolian housing, they use coal as a fuel for indoor heating and cooking. The combustion of coal releases particulate matter, which can affect indoor air quality. The purpose of this study was to analyze the concentrations of indoor $PM_{2.5}$ in winter time dwellings in ger town. Methods: In this study, indoor $PM_{2.5}$ concentrations, temperature and humidity in houses were measured by a real-time PM monitor, while the time activity patterns of the residents were also observed. Results: The correlation between factors that may affect the indoor air quality was analyzed.The indoor $PM_{2.5}$ concentrations were $178.4{\pm}152.7{\mu}g/m^3$ (n=37). Five types of indoor $PM_{2.5}$ concentrations have been classified, which were associated with indoor activity. The stove type, fuel types and indoor activities such as cleaning, cooking and opening the stoves were not significantly associated with indoor $PM_{2.5}$ levels. Conclusions: Further study is needed to determine the effect of stove type through 24hours of indoor air quality monitoring.

• Journal of Korean Society of Transportation
• /
• v.26 no.4
• /
• pp.251-264
• /
• 2008

One of the crucial elements to fully facilitate the various benefits of intelligent transportation systems (ITS) is to obtain more reliable traffic monitoring in real time. To date, point and section-based traffic measurements have been available through existing surveillance technologies, such as loops and automatic vehicle identification (AVI) systems. However, seamless and more reliable traffic data are required for more effective traffic information provision and operations. Technology advancements including vehicle tracking and wireless communication enable the acceleration of the availability of individual vehicle travel information. This study presents a UBIquitous PRObe vehicle Surveillance System (UBIPROSS) using vehicle-to-vehicle (V2V) wireless communications. Seamless vehicle travel information, including origin-destination information, speed, travel times, and other data, can be obtained by the proposed UBIPROSS. A set of parameters associated with functional requirements of the UBIPROSS, which include the market penetration rate (MPR) of equipped vehicles, V2V communication range, and travel time update interval, are investigated by a Monte Carlo simulation- (MCS) based evaluation framework. In addition, this paper describes prototypical implementation. Field test results and identified technical issues are also discussed. It is expected that the proposed system would be an invaluable precursor to develop a next-generation traffic surveillance system.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.10 no.3
• /
• pp.165-175
• /
• 2006

To analyze the deformation and failure of slopes, generally, two types of model, Polynomial model and Growth model, are applied. These two models are focused on the behavior of the slope by time. Therefore, this research is more focused on predicting of slope failure than analyzing the slope behavior by time. Generally, Growth model is used to analyze the soil slope, to the contrary, Polynomial model is used for rock slope. However, 3-degree polynomial($y=ax^3+bx^2+cx+d$) is suggested to combine two models in this research. The main trait of this model is having an asymptote. The fields to adopt this model are Gosujae Danyang(soil slope) and Youngduk slope(rock slope), which are the cut-slope near national road. Data from Gosujae are shown the failure traits of soil slope, to the contrary, those of Youngduk slope are shown the traits of rock slope. From the real-time monitoring data of the slope, 3-degree polynomial is proved as excellent system to analyze the failure and behavior of slope. In case of Polynomial model, even if the order of polynomials is increased, the $R^2$ value and shape of the curve-fitted graph is almost the same.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.21 no.5
• /
• pp.103-111
• /
• 2021

The purpose of this study is to study a method for preventing collisions of multiple drones in controlled airspace. As a result of the study, it was proved that it is appropriate as a method to control drone collisions after setting accurate information on the ROI (Region of Interest) area estimated based on the expected drone path and time in the control system as a method to avoid drone collision. As a result of the empirical analysis, the diameter of the flight path of the operating drone should be selected to reduce the risk of collision, and the change in the departure time and operating speed of the operating drone did not act as an influencing factor in the collision. In addition, it has been demonstrated that providing flight priority is one of the appropriate methods as a countermeasure to avoid collisions. For collision avoidance methods, not only drone sensor-based collision avoidance, but also collision avoidance can be doubled by monitoring and predicting collisions in the control system and performing real-time control. This study is meaningful in that it provided an idea for a method for preventing collisions of multiple drones in controlled airspace and conducted practical tests. This helps to solve the problem of collisions that occur when multiple drones of different types are operating based on the control system. This study will contribute to the development of related industries by preventing accidents caused by drone collisions and providing a safe drone operation environment.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.3
• /
• pp.527-535
• /
• 2021

Traditional business intelligence (BI) systems have been used widely as tools for better decision-making on time. On the other hand, building a data warehouse (DW) for the efficient analysis of rapidly growing data is time-consuming and complex. In particular, the ETL (Extract, Transform, and Load) process required to build a data warehouse has become much more complex as the BI platform moves to a cloud environment. Various BI solutions based on the NoSQL database, such as MongoDB, have been proposed to overcome these ETL issues. Decision-makers want easy access to data without the help of IT departments or BI experts. Recently, self-service BI (SSBI) has emerged as a way to solve these BI issues. This paper proposes a self-service BI system with farming data using the MongoDB cloud as DW to support the selection of new crops by return-farmers. The proposed system includes functions to provide insights to decision-makers, including data visualization using MongoDB charts, reporting for advanced data search, and monitoring for real-time data analysis. Decision makers can access data directly in various ways and can analyze data in a self-service method using the functions of the proposed system.

• Journal of Environmental Analysis, Health and Toxicology
• /
• v.21 no.4
• /
• pp.264-280
• /
• 2018

In this study, volatile organic compounds (VOCs) were measured using a proton transfer reaction-time of flight-mass spectrometer (PTR-ToF-MS) at the Seoul Metropolitan Area Intensive Monitoring Station (SIMS) in Korea during the summer season of 2018. The results revealed that oxygenated VOCs (OVOCs) contributed a large fraction (83.6%) of the total VOCs, with methanol being the most abundant constituent (38.6%). The VOCs measured at SIMS were strongly influenced by local conditions. Non-volatile organic compounds (NVOCs), such as pinene, increased due to northeasterly wind direction in the morning, and OVOCs and anthropogenic VOCS (AVOCs) increased with northwesterly wind direction during the daytime. This was the result of the eastward location of Bukhansan National Park and the westward location of urban area from the SIMS location. The VOCs included abundant oxidized forms of VOCs, which can affect the generation of fine dust through various response pathways in the atmosphere. The real-time measurement technique using PTR-ToF-MS suggested in this study is expected to contribute to an improved scientific understanding of high-concentration fine dust events because the high temporal resolution makes it possible to analyze the variations of VOCs reflected in dynamic events.