• Korean Journal of Remote Sensing
• /
• v.36 no.1
• /
• pp.29-39
• /
• 2020

Forest fire is a critical disaster that causes massive destruction of forest ecosystem and economic loss. Hence, accurate estimation of the burned area is important for evaluation of the degree of damage and for preparing baseline data for recovery. Since most of the area size damaged by wildfires in Korea is less than 1 ha, it is necessary to use satellite or drone images with a resolution of less than 10m for detecting the damage area. This paper aims to detect wildfire-damaged area from a Kompsat-3 image using the indices such as NDVI (normalized difference vegetation index) and FBI (fire burn index) and to examine the classification characteristics according to the methods such as Otsu thresholding and ISODATA(iterative self-organizing data analysis technique). To mitigate the salt-and-pepper phenomenon of the pixel-based classification, a gaussian filter was applied to the images of NDVI and FBI. Otsu thresholding and ISODATA could distinguish the burned forest from normal forest appropriately, and the salt-and-pepper phenomenon at the boundaries of burned forest was reduced by the gaussian filter. The result from ISODATA with gaussian filter using NDVI was closest to the official record of damage area (56.9 ha) published by the Korea Forest Service. Unlike Otsu thresholding for binary classification,since the ISODATA categorizes the images into multiple classes such as(1)severely burned area, (2) moderately burned area, (3) mixture of burned and unburned areas, and (4) unburned area, the characteristics of the boundaries consisting of burned and normal forests can be better expressed. It is expected that our approach can be utilized for the high-resolution images obtained from other satellites and drones.

• Korean Journal of Remote Sensing
• /
• v.37 no.5_3
• /
• pp.1373-1387
• /
• 2021

Forest fire poses a significant threat to the environment and society, affecting carbon cycle and surface energy balance, and resulting in socioeconomic losses. Widely used multi-spectral satellite image-based approaches for burned area detection have a problem in that they do not work under cloudy conditions. Therefore, in this study, Sentinel-1 Synthetic Aperture Radar (SAR) data from Europe Space Agency, which can be collected in all weather conditions, were used to identify forest fire damaged area based on a series of processes including Principal Component Analysis (PCA) and K-means clustering. Four forest fire cases, which occurred in Gangneung·Donghae and Goseong·Sokcho in Gangwon-do of South Korea and two areas in North Korea on April 4, 2019, were examined. The estimated burned areas were evaluated using fire reference data provided by the National Institute of Forest Science (NIFOS) for two forest fire cases in South Korea, and differenced normalized burn ratio (dNBR) for all four cases. The average accuracy using the NIFOS reference data was 86% for the Gangneung·Donghae and Goseong·Sokcho fires. Evaluation using dNBR showed an average accuracy of 84% for all four forest fire cases. It was also confirmed that the stronger the burned intensity, the higher detection the accuracy, and vice versa. Given the advantage of SAR remote sensing, the proposed statistical processing and K-means clustering-based approach can be used to quickly identify forest fire damaged area across the Korean Peninsula, where a cloud cover rate is high and small-scale forest fires frequently occur.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.38 no.3
• /
• pp.187-196
• /
• 2020

In this study, collapsed building detection using UAV (Unmanned Aerial Vehicle) and PlanetScope satellite images was carried out, suggesting the possibility of utilization of heterogeneous sensors in object detection located on the surface. To this end, the area where about 20 buildings collapsed due to forest fire damage was selected as study site. First of all, the feature information of objects such as ExG (Excess Green), GLCM (Gray-Level Co-Occurrence Matrix), and DSM (Digital Surface Model) were generated using high-resolution UAV images performed object-based segmentation to detect collapsed buildings. The features were then used to detect candidates for collapsed buildings. In this process, a result of the change detection using PlanetScope were used together to improve detection accuracy. More specifically, the changed pixels acquired by the bitemporal PlanetScope images were used as seed pixels to correct the misdetected and overdetected areas in the candidate group of collapsed buildings. The accuracy of the detection results of collapse buildings using only UAV image and the accuracy of collapse building detection result when UAV and PlanetScope images were used together were analyzed through the manually dizitized reference image. As a result, the results using only UAV image had 0.4867 F1-score, and the results using UAV and PlanetScope images together showed that the value improved to 0.8064 F1-score. Moreover, the Kappa coefficiant value was also dramatically improved from 0.3674 to 0.8225.

• Journal of Digital Contents Society
• /
• v.18 no.1
• /
• pp.71-77
• /
• 2017

Virtual advertisement is an advertising technique that using computer graphic in a media production such as a sports image for inserting product image, logo, advertising slogan, etc. Recently, the image insertion of virtual advertisement is actively spreading due to the satisfaction of technical element for the image insertion of virtual advertisement in sports advertisement by increasing of the image processing technology and the computing performance. In addition, image processing technology for automatic insertion has become an important research field in the virtual advertisement field. In this paper, we propose the method of extracting less-informative region by using image processing technique and machine learning to insert a virtual advertisement automatically in sports image. The proposed method analyzes the brightness level of image through the histogram and extracts the less-informative region using the machine learning method.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.49 no.1
• /
• pp.63-73
• /
• 2021

This paper presents a algorithm for automatic target recognition robust to the influence of the flame in order to track the target by EOTS(Electro-Optical Targeting System) equipped on UAV(Unmanned Aerial Vehicle) when there is aerial target or marine target with flame at the same time. The proposed method converts infrared images of targets and flames into a gradient vector field, and applies each gradient magnitude to a polynomial curve fitting technique to extract polynomial coefficients, and learns them in a shallow neural network model to automatically recognize targets and flames. The performance of the proposed technique was confirmed by utilizing the various infrared image database of the target and flame. Using this algorithm, it can be applied to areas where collision avoidance, forest fire detection, automatic detection and recognition of targets in the air and sea during automatic flight of unmanned aircraft.

• Convergence Security Journal
• /
• v.21 no.1
• /
• pp.191-200
• /
• 2021

Artificial intelligence technology is developing in an environment where a lot of data is produced due to the development of computing technology, a cloud environment that can store data, and the spread of personal mobile phones. Among these artificial intelligence technologies, the deep neural network provides excellent performance in image recognition and image classification. There have been many studies on image detection for forest fires and fire prevention using such a deep neural network, but studies on detection of cigarette smoking were insufficient. Meanwhile, military units are establishing surveillance systems for various facilities through CCTV, and it is necessary to detect smoking near ammunition stores or non-smoking areas to prevent fires and explosions. In this paper, by reflecting experimentally optimized numerical values such as activation function and learning rate, we did the detection of smoking pictures and non-smoking pictures in two cases. As experimental data, data was constructed by crawling using pictures of smoking and non-smoking published on the Internet, and a machine learning library was used. As a result of the experiment, when the learning rate is 0.004 and the optimization algorithm Adam is used, it can be seen that the accuracy of 93% and F1-score of 94% are obtained.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.21 no.1
• /
• pp.47-60
• /
• 2018

In this paper, we proposed a method for eliminating aiming error of unguided anti-tank rocket using improved target tracking. Since predicted fire is necessary to hit moving targets with unguided rockets, a method was proposed to estimate the position and velocity of target using fire control system. However, such a method has a problem that the hit rate may be lowered due to the aiming error of the shooter. In order to solve this problem, we used an image-based target tracking method to correct error caused by the shooter. We also proposed a robust tracking method based on TLD(Tracking Learning Detection) considering characteristics of the FCS(Fire Control System) devices. To verify the performance of our proposed algorithm, we measured the target velocity using GPS and compared it with our estimation. It is proved that our method is robust to shooter's aiming error.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.21 no.1
• /
• pp.61-68
• /
• 2021

In the past, radars have been classified into fire control radars, detection radars, tracking radars, and image acquisition radars according to the characteristics of the mission. However, multi-function radars perform various tasks within a single system, such as target detection, tracking, identification friend or foe, jammer detection and response. Therefore, efficient resource management is essential to operate multi-function radars with limited resources. In particular, the target threat for tracking the detected target and the method of selecting the tracking cycle based on this is an important issue. If focus on tracking a threat target, Radar can't efficiently manage the targets detected in other areas, and if you focus on detection, tracking performance may decrease. Therefore, effective scheduling is essential. In this paper, we propose the TaP (Time and Priority) algorithm, which is a multi-functional radar scheduling scheme, and a software design method to construct it.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.21 no.3
• /
• pp.419-432
• /
• 2019

Most of deep learning model training was proceeded by supervised learning, which is to train labeling data composed by inputs and corresponding outputs. Labeling data was directly generated manually, so labeling accuracy of data is relatively high. However, it requires heavy efforts in securing data because of cost and time. Additionally, the main goal of supervised learning is to improve detection performance for 'True Positive' data but not to reduce occurrence of 'False Positive' data. In this paper, the occurrence of unpredictable 'False Positive' appears by trained modes with labeling data and 'True Positive' data in monitoring of deep learning-based CCTV accident detection system, which is under operation at a tunnel monitoring center. Those types of 'False Positive' to 'fire' or 'person' objects were frequently taking place for lights of working vehicle, reflecting sunlight at tunnel entrance, long black feature which occurs to the part of lane or car, etc. To solve this problem, a deep learning model was developed by simultaneously training the 'False Positive' data generated in the field and the labeling data. As a result, in comparison with the model that was trained only by the existing labeling data, the re-inference performance with respect to the labeling data was improved. In addition, re-inference of the 'False Positive' data shows that the number of 'False Positive' for the persons were more reduced in case of training model including many 'False Positive' data. By training of the 'False Positive' data, the capability of field application of the deep learning model was improved automatically.

• Journal of Internet Computing and Services
• /
• v.23 no.4
• /
• pp.57-64
• /
• 2022

If the inside of a building collapses due to a disaster such as fire, collapse, or natural disaster, the physical security inside the building is likely to become ineffective. Here, physical security is needed to minimize the human casualties and physical damages in the collapsed building. Therefore, this paper proposes an algorithm to minimize the damage in a disaster situation by fusing existing research that detects obstacles and collapsed areas in the building and a deep learning-based object detection algorithm that minimizes human casualties. The existing research uses a single camera to determine whether the corridor environment in which the robot is currently located has collapsed and detects obstacles that interfere with the search and rescue operation. Here, objects inside the collapsed building have irregular shapes due to the debris or collapse of the building, and they are classified and detected as obstacles. We also propose a method to detect rescue requesters-the most important resource in the disaster situation-and minimize human casualties. To this end, we collected open-source disaster images and image data of disaster situations and calculated the accuracy of detecting rescue requesters in disaster situations through various deep learning-based object detection algorithms. In this study, as a result of analyzing the algorithms that detect rescue requesters in disaster situations, we have found that the YOLOv4 algorithm has an accuracy of 0.94, proving that it is most suitable for use in actual disaster situations. This paper will be helpful for performing efficient search and rescue in disaster situations and achieving a high level of physical security, even in collapsed buildings.