• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.23 no.1
• /
• pp.65-70
• /
• 2023

In this paper, we used a deep learning model to detect and segment flame and smoke in real time from fires. To this end, well known U-NET was used to separate and divide the flame and smoke of the fire using multi-class. As a result of learning using the proposed technique, the values of loss error and accuracy are very good at 0.0486 and 0.97996, respectively. The IOU value used in object detection is also very good at 0.849. As a result of predicting fire images that were not used for learning using the learned model, the flame and smoke of fire are well detected and segmented, and smoke color were well distinguished. Proposed method can be used to build fire prediction and detection system.

• Journal of Multimedia Information System
• /
• v.9 no.3
• /
• pp.233-244
• /
• 2022

A MATLAB program was developed to calculate the half-wavelength of a sine-curve baseband signal with white noise by using an autocorrelation function, a SG filter, and zero-crossing detection. The frequency of the input signal can be estimated from 1) the first zero-crossing (corresponding to ¼λ) and 2) the R value (the Y axis of the correlogram) at the center of the segment. Thereby, the frequency information of the preceding segment can be obtained. If the segment size were optimized, and a portion with a large zero-crossing dynamic range were obtained, the frequency discrimination ability would improve. Furthermore, if the values of the correlogram for each frequency prepared on the CPU side were prepared in a table, the volume of calculations can be reduced by 98%. As background, period detection by autocorrelation coefficients requires an integer multiple of 1/2λ (when using a sine wave as the object of the autocorrelation function), otherwise the correlogram drawn by R value will not exhibit orthogonality. Therefore, it has not been used in bio-telemetry where the frequencies move around.

• The Journal of Korea Robotics Society
• /
• v.17 no.3
• /
• pp.245-254
• /
• 2022

Sensor dataset for autonomous driving is one of the essential components as the deep learning approaches are widely used. However, most driving datasets are focused on typical environments such as sunny or cloudy. In addition, most datasets deal with color images and lidar. In this paper, we propose a driving dataset with multi-spectral images and lidar in adverse weather conditions such as snowy, rainy, smoky, and dusty. The proposed data acquisition system has 4 types of cameras (color, near-infrared, shortwave, thermal), 1 lidar, 2 radars, and a navigation sensor. Our dataset is the first dataset that handles multi-spectral cameras in adverse weather conditions. The Proposed dataset is annotated as 2D semantic labels, 3D semantic labels, and 2D/3D bounding boxes. Many tasks are available on our dataset, for example, object detection and driveable region detection. We also present some experimental results on the adverse weather dataset.

• Journal of KIISE
• /
• v.43 no.7
• /
• pp.795-800
• /
• 2016

In image classification, recent CNNs compete with human performance. However, there are limitations in more general recognition. Herein we deal with indoor images that contain too much information to be directly processed and require information reduction before recognition. To reduce the amount of data processing, typically variational inference or variational Bayesian methods are suggested for object detection. However, these methods suffer from the difficulty of marginalizing over the given space. In this study, we propose an image-text integrated recognition system using active vision based on Spatial Transformer Networks. The system attempts to efficiently sample a partial region of a given image for a given language information. Our experimental results demonstrate a significant improvement over traditional approaches. We also discuss the results of qualitative analysis of sampled images, model characteristics, and its limitations.

• Smart Media Journal
• /
• v.12 no.11
• /
• pp.57-66
• /
• 2023

In this paper, we present a study on an AI-based system utilizing the CCTV system within city buses to predict the intentions of boarding and alighting passengers, with the aim of preventing accidents. The proposed system employs the YOLOv7 Pose model to detect passengers, while utilizing an LSTM model to predict intentions of tracked passengers. The system can be installed on the bus's CCTV terminals, allowing for real-time visual confirmation of passengers' intentions throughout driving. It also provides alerts to the driver, mitigating potential accidents during passenger transitions. Test results show accuracy rates of 0.81 for analyzing boarding intentions and 0.79 for predicting alighting intentions onboard. To ensure real-time performance, we verified that a minimum of 5 frames per second analysis is achievable in a GPU environment. his algorithm enhance the safety of passenger transitions during bus operations. In the future, with improved hardware specifications and abundant data collection, the system's expansion into various safety-related metrics is promising. This algorithm is anticipated to play a pivotal role in ensuring safety when autonomous driving becomes commercialized. Additionally, its applicability could extend to other modes of public transportation, such as subways and all forms of mass transit, contributing to the overall safety of public transportation systems.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.4 no.1
• /
• pp.267-282
• /
• 2000

With the development of computer&network technology and the growth of its dependance, computer failures not only lose human and material resources but also make organization's competition weak as a side-effect of information society. Therefore, people consider computer security as important factor. Intrusion Detection Systems (IDS) detect intrusions and take an appropriate action against them in order to protect a computer from system failure due to illegal intrusion. A variety of methods and models for IDS have been developed until now, but the existing methods or models aren't enough to detect intrusions because of the complexity of computer network the vulnerability of the object system, insufficient understanding for information security and the appearance of new illegal intrusion method. We propose a new IDS model to be suitable at the information system organized by homogeneous hosts and design for the IDS model and implement the prototype of it for feasibility study. The IDS model consist of many distributed unit sensor IDSs at homogeneous hosts and if any of distributed unit sensor IDSs detect anomaly system call among system call sequences generated by a process, the anomaly system call can be dynamically shared with other unit sensor IDSs. This makes the IDS model can effectively detect new intruders about whole information system.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.29 no.5
• /
• pp.421-426
• /
• 2009

The emissivity is different because the emitted angle changes according to the position of the infrared thermography camera and object. Because of this, the temperature distribution expressed when measuring the temperature by using the infrared thermography system is not the accuracy temperature. Although the real surface temperature is constant, the temperature measured by using infrared thermography camera have error in accordance with the value of emissivity. In this paper, the temperatures of the round cylindrical object and the flat square object that heated to the equal temperature were measured by infrared thermography camera. The emissivity calibration formula and correction table are made with the affect of the view angle and emission angle form the surface temperature value. The error of measured temperature values are corrected by using the emissivity calibration formula and correction table, and apply to defect detection of the nuclear power plant pipe. From the calibration method, reliability surface temperature values were obtained.

• Korean Journal of Remote Sensing
• /
• v.27 no.4
• /
• pp.421-433
• /
• 2011

The importance of acquisition of road information has recently been increased with a rapid growth of spatial-related services such as urban information system and location based service. This paper proposes an automatic road extraction method using object-based approach which was issued alternative of pixel-based method recently. Firstly, the spatial objects were created by MSRS(Modified Seeded Region Growing) method, and then the key road objects were extracted by using properties of objects such as their shape feature information and adjacency. The omitted road objects were also traced considering spatial correlation between extracted road and their neighboring objects. In the end, the final road region was extracted by connecting discontinuous road sections and improving road surfaces through their geometric properties. To assess the proposed method, quantitative analysis was carried out. From the experiments, the proposed method generally showed high road detection accuracy and had a great potential for the road extraction from high resolution satellite images.

• Journal of Urban Science
• /
• v.7 no.2
• /
• pp.53-60
• /
• 2018

Digital Twin is a technology that creates a photocopy of real-world objects on a computer and analyzes the past and present operational status by fusing the structure, context, and operation of various physical systems with property information, and predicts the future society's countermeasures. In particular, 3D rendering technology (UAS, LiDAR, GNSS, etc.) is a core technology in digital twin. so, the research and application are actively performed in the industry in recent years. However, UAS (Unmanned Aerial System) and LiDAR (Light Detection And Ranging) have to be solved by compensating blind spot which is not reconstructed according to the object shape. In addition, the terrestrial LiDAR can acquire the point cloud of the object more precisely and quickly at a short distance, but a blind spot is generated at the upper part of the object, thereby imposing restrictions on the forward digital twin modeling. The UAS is capable of modeling a specific range of objects with high accuracy by using high resolution images at low altitudes, and has the advantage of generating a high density point group based on SfM (Structure-from-Motion) image analysis technology. However, It is relatively far from the target LiDAR than the terrestrial LiDAR, and it takes time to analyze the image. In particular, it is necessary to reduce the accuracy of the side part and compensate the blind spot. By re-optimizing it after fusion with UAS and Terrestrial LiDAR, the residual error of each modeling method was compensated and the mutual correction result was obtained. The accuracy of fusion-based 3D model is less than 1cm and it is expected to be useful for digital twin construction.

• Journal of Sensor Science and Technology
• /
• v.19 no.1
• /
• pp.8-16
• /
• 2010

This research proposes measuring of BCG(ballistocardiogram) to monitor heart activities in a non-constrained environment, at home or work. Unlike with ECG, measuring BCG does not require the attachment of leads on the subject's body and allows signal measuring in a non-constrained state. It enables effective long-term monitoring of cardiac conditions. In this study a chair type BCG measurement system to continuous monitor the activity of the heart is implemented. The instrument consists of upper petal and ready for press of chair load cell sensor is attached to measure the change of the object's weight. In order to extract the output ballistic signal from the weight and force sensor signals. Beside the signal processing circuit for the digital conversion, the ballistic signal is detected using DAQ equipment. Signal processing algorithm including wavelet transforms for noise cancellation, template matching for normalization and peak detection in BCG is developed. ECG and BCG were concurrently measured to evaluate the performance of the system, and comparing the characteristics of the two signals verified the possibility of the system in non-constrained and nonconscious health monitoring.