• 한국항행학회논문지
• /
• 제27권5호
• /
• pp.685-689
• /
• 2023

The proposed system in the study aims to detect forest fires in real-time stream data received from the drone-camera. Recently, the number of wildfires has been increasing, and also the large scaled wildfires are frequent more and more. In order to prevent forest fire damage, many experiments using the drone camera and vision analysis are actively conducted, however there were many challenges, such as network speed, pre-processing, and model performance, to detect forest fires from real-time streaming data of the flying drone. Therefore, this study applied image data processing works to capture five good image frames for vision analysis from whole streaming data and then developed the object detection model based on YOLO_v2. As the result, the classification model performance of forest fire images reached upto 93% of accuracy, and the field test for the model verification detected the forest fire with about 70% accuracy.

• 로봇학회논문지
• /
• 제12권4호
• /
• pp.425-431
• /
• 2017

As drones gain more popularity these days, drone detection becomes more important part of the drone systems for safety, privacy, crime prevention and etc. However, existing drone detection systems are expensive and heavy so that they are only suitable for industrial or military purpose. This paper proposes a novel approach for training Convolutional Neural Networks to detect drones from images that can be used in embedded systems. Unlike previous works that consider the class probability of the image areas where the class object exists, the proposed approach takes account of all areas in the image for robust classification and object detection. Moreover, a novel loss function is proposed for the CNN to learn more effectively from limited amount of training data. The experimental results with various drone images show that the proposed approach performs efficiently in real drone detection scenarios.

• Corrosion Science and Technology
• /
• 제18권6호
• /
• pp.253-257
• /
• 2019

A deep learning based real-time painting surface inspection algorithm is proposed herein, designed for developing an autonomous inspection drone. The painting surface inspection is usually conducted manually. However, the manual inspection has a limitation in obtaining accurate data for correct judgement on the surface because of human error and deviation of individual inspection experiences. The best method to replace manual surface inspection is the vision-based inspection method with a camera, using various image processing algorithms. Nevertheless, the visual inspection is difficult to apply to surface inspection due to diverse appearances of material, hue, and lightning effects. To overcome technical limitations, a deep learning-based pattern recognition algorithm is proposed, which is specialized for painting surface inspections. The proposed algorithm functions in real time on the embedded board mounted on an autonomous inspection drone. The inspection results data are stored in the database and used for training the deep learning algorithm to improve performance. The various experiments for pre-inspection of painting processes are performed to verify real-time performance of the proposed deep learning algorithm.

• 대한임베디드공학회논문지
• /
• 제18권2호
• /
• pp.75-80
• /
• 2023

Unmanned drone stations for automatic charging have been developed in order to overcome the flying time limitation of rotary wing drones. Since the drone stations is an unmanned operating system, each of the drones will be required to have a high degree of landing accuracy. Drone precision landing has been mainly studied depended on image processing technologies, but the image processing systems make several problems, such as the mission weight, the drone cost, and the development complexity increases, and the flight time decrease. Thus, this paper researched accuracy of precision landing based on RTK (real time kinetics) for rotary wing drones. For the experiments of RTK based precision landing, a drone repeatedly performed three missions. The survey accuracies of the RTK about missions respectively were set as 0.3, 0.2, and 0.1 meters. Each mission has one take-off point, two way-points and one landing-point, and was repeated ten times. The experiment results revealed landing error distance means of around 0.258, 0.12 and 0.057 meters on each of RTK setting.

• 융합정보논문지
• /
• 제9권12호
• /
• pp.17-23
• /
• 2019

본 논문에서는 미니 드론을 조종하면서 드론에 부착된 카메라가 촬영하는 영상을 실시간으로 받아들여 특정인의 얼굴을 인식하여 확인시켜주는 시스템 개발 방법론을 제안한다. 본 시스템의 개발을 위해서는 OpenCV, Python 관련 라이브러리 및 드론 SDK 등을 사용한다. 실시간 드론 영상으로부터 특정인의 얼굴 인식 비율을 높이기 위해서는 딥러닝 기반의 얼굴 인식 알고리즘을 사용하며 특히 Triples 원리를 활용한다. 시스템의 성능을 확인하기 위해 저자 얼굴을 기준으로 30회 동안 얼굴 인식 실험을 수행한 결과 약 95% 이상의 인식률을 보여주었다. 본 논문의 연구 결과물은 관광지, 축제 행사장 등에서 특정인을 드론으로 빠르게 찾기 위한 목적으로 사용할 수 있을 것으로 판단된다.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2018년도 학술발표회
• /
• pp.30-30
• /
• 2018

제방 매핑 연구는 제방 안전도 측정을 위해 중요하며, 드론 영상은 실시간 제방 지도를 만들기 위해 활용될 수 있다. 본 연구에서는 하천유역을 촬영한 드론 영상을 이용하여 제방 매핑 연구를 진행하였다. 우선 테스트 베드의 하천유역을 촬영한 드론 영상을 수집한다. 그리고 드론 영상을 이용하여 사진측량/영상처리 기법을 기반으로 3차원 디지털 표고모델(DEM)과 정사영상을 제작한다. 최종적으로 수동/자동화 기법을 이용하여 제방 표면에 존재하는 중요한 객체(둑마루, 제방법선, 비탈경사면, 침식지역 등)를 디지털 표고모델과 정사영상으로부터 탐지한다. 추후 연구에서는 드론 영상으로부터 중요한 객체를 탐지할 수 있는 자동화 기술 개발을 목표로 한다.

• 전기학회논문지
• /
• 제67권6호
• /
• pp.773-781
• /
• 2018

In recent years, a study of power-line inspection using an unmanned aerial vehicle (UAV) has been actively conducted. However, relevant studies have been conducting power-line inspection with an UAV operated by manual control, and they have developed just power-line detection algorithm on aerial images. To overcome limitations of existing research, we propose a drone-based power-line tracking system in this paper. The main contributions of this paper are to operate developed system under configured environment and to develop a power-line detection algorithm in real-time. Developed system is composed of the power-line detection and the image-based tracking control. To detect a power-line in real-time, a region of interest (ROI) image is extracted. Furthermore, clustering algorithm is used in order to discriminate the power-line from background. Finally, the power-line is detected by using the Hough transform, and a center position and a tilt angle are estimated by using the Kalman filter to control a drone smoothly. We design a position controller and an attitude controller for image-based tracking control, and both controllers are designed based on the proportional-derivative (PD) control method. The interaction between the position controller and the attitude controller makes the drone track the power-line. Several experiments were carried out in environments where conditions are similar to actual environments, which demonstrates the superiority of the developed system.

• International Journal of Advanced Culture Technology
• /
• 제11권2호
• /
• pp.381-388
• /
• 2023

Small-scale environmental impact assessments have limitations in terms of survey duration and evaluation resources, which can hinder the assessment and analysis of the current situation. In this study, we propose the use of drone technology during the environmental impact assessment process to supplement these limitations in the current situation analysis. Drone photography can provide rapid and accurate high-resolution images, allowing for the collection of various information about the target area. This information can include different types of data such as terrain, vegetation, landscape, and real-time 3D spatial information, which can be collected and processed using GIS software to understand and analyze the environmental conditions. In this study, we confirmed that terrain and vegetation analysis and prediction of the target area using drone photography and GIS analysis software is possible, providing useful information for environmental impact assessments.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2021년도 춘계학술대회
• /
• pp.538-540
• /
• 2021

본 논문에서는 드론의 표적탐지 임무 수행 시 상대운동 정보 제공을 위하여 지정된 표적을 탐지하고 인식하는 스마트 표적탐지 시스템을 제안하였다. 제안된 시스템은 적절한 정확도(i.e. mAP, IoU) 및 높은 실시간성을 동시에 확보할 수 있는 알고리즘을 개발하는데 중점을 두었다. 제안된 시스템은 Google Inception V2 딥러닝 모델의 100k 학습 후 test 결과가 1.0에 가까운 정확성을 보였고 실시간성도 Nvidia GTX 2070 Max-Q를 기반으로 한 고성능 노트북 활용 시에 추론 속도가 약 60-80[Hz]를 기록하였다. 제안된 스마트 표적탐지 시스템은 드론과 같이 운용되어 컴퓨터 영상처리를 활용하여 표적을 자동으로 인식하고 표적을 따라가면서 감시정찰 임무를 성공적으로 수행하는데 도움이 될 것이다.

• 한국산업융합학회 논문집
• /
• 제26권5호
• /
• pp.775-784
• /
• 2023

Drone images are being used more and more actively in the fields of surveying and spatial information, and are rapidly replacing existing aerial and satellite images. The technology of quickly acquiring real-time data at low cost and processing it is now being applied to actual industries beyond research. However, there are also problems encountered as this progresses. When high-resolution spatial information is acquired using a general 2D flight plan for a terrain with sever undulations, problems arise due to the difference in resolution of the data. In particular, when a low-altitude high-resolution image is taken using a drone in a mountainous or steep terrain, there may be a problem in image matching due to a resolution difference caused by terrain undulations. This problem occurs because a drone acquires data while flying on a 2D plane at a fixed altitude, just like conventional aerial photography. In order to acquire high-quality 3D data using a drone, the scale difference for the shooting distance should be considered. In addition, in order to obtain facade images of large structures, it is necessary to take images in 3D space. In this study, in order to improve the disadvantages of the 2D flight method, a 3D flight plan was established for the study area, and it was confirmed that high-quality 3D spatial information could be obtained in this way.