http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
정우진 ( Woojin Chung ),장순웅 ( Soonwoong Chang ) 한국환경과학회 2016 한국환경과학회지 Vol.25 No.7
Environmental policy implementation has been strengthened to protect the source waters in Korea and to improve their water quality. Increasing of non-point source caused water quality problem continuously. Research on runoff from forests, which occupy over 65% of the land in korea, is insufficient, and studies on the characteristics and influences of storm runoff are necessary. In this study, we chose to compare the effects of land use in the form of two types of forest distribution and then gathered data on storm characteristics and runoff properties during rainfall events in these areas. Furthermore, the significance and influences of the discharges were analyzed through correlation analysis, and multilateral runoff characteristics were examined by deducing a formula through COD(Mn) and TOC regression analysis. At two forest points, for which the basin areas differed from each other, flow changed according to storm quantity and intensity. The peak discharge at point A, where the basin area was big, was high, whereas water-quality fundamental items (BOD, COD(Mn), and SS) and TOC density were high at point B where the slope and storm intensity were high. Effects of dissolved organic matter were determined through correlation analysis, and the regression formulas for COD(Mn) and TOC were deduced by regression analysis. It is expected that the data from this study could be useful as basic information in establishing forest management measures.
딥러닝 기반 차량 내 블랙박스를 이용한 실시간 도로 위 정지 차량 검지 알고리즘 개발
정우진(Woojin Jeong),이재용(Jaeyong Lee),김창일(Chang-il Kim),박진욱(Jinuk Park),박용주(Yongju Park) 대한전자공학회 2023 대한전자공학회 학술대회 Vol.2023 No.6
In this study, we propose a method for detecting a halted vehicle from a dashcam footage in a moving vehicle. Identifying a halted vehicle has several challenges such as camera lens curvature, perspective distortions, changing object positions based on velocity and distance, and variable road conditions. To address these challenges, we propose a Region of Interest (RoI) within the dashcam video stream, extracting optical flow values based on the Gunnar Farneback algorithm. Additionally, we employ YOLOv5 and DeepSORT tracking algorithms to identify the object area. By comparing the flow within and surrounding the object area, we can discern whether the vehicle is stationary. The proposed method provides a robust approach for the detection of stopped vehicles in dynamic and complex traffic environments.
객체 인식 모델을 활용한 적재불량 화물차 탐지 시스템 개발
정우진(Woojin Jung),박용주(Yongju Park),박진욱(Jinuk Park),김창일(Chang-il Kim) 한국정보통신학회 2022 한국정보통신학회 종합학술대회 논문집 Vol.26 No.2
최근 증가하고 있는 도로 위 적재 불량 화물차는 비정상적인 무게 중심으로 인해 물체 낙하, 도로파손, 연쇄 추돌 등 교통 안전에 위해가 되고 한번 사고가 발생하면 큰 피해가 유발할 수 있다. 하지만 이러한 비정상적인 무게 중심은 적재 불량 차량 인식을 위한 주행 중 축중 시스템으로는 검출이 불가능하다는 한계점이 있다. 본 논문에서는 이러한 사회 문제를 야기하는 적재 불량 차량을 관리하기 위한 객체 인식 기반 AI 모델을 구축하고자 한다. 또한 AI-Hub에 공개된 약 40만장의 대형차, 소형차, 중형차 별 적재 불량 차량과 일반차량으로 구분 된 데이터 셋 중 종류별로 제공되는 CCTV, 블랙박스, 카메라 시점의 적재 불량 차량 데이터 셋을 분석하여 전처리를 통해 적재 불량 차량 검지 AI 모델의 성능을 향상시키는 방법을 제시한다. 이를 통해, 원시 데이터를 활용한 학습 성능 대비 약 23% 향상된 적재 불량 차량의 검출 성능을 나타냄을 보였다. 본 연구 결과를 통해 공개 빅데이터를 보다 효율적으로 활용하여, 객체 인식 기반 적재 불량 차량 탐지 모델 개발에 적용할 수 있을 것으로 기대된다. Recently, the increasing number of overloaded vehicles on the road poses a risk to traffic safety, such as falling objects, road damage, and chain collisions due to the abnormal weight distribution, and can cause great damage once an accident occurs. However, this irregular weight distribution is not possible to be recognized with the current weight measurement system for vehicles on roads. To address this limitation, we propose to build an object detection-based AI model to identify overloaded vehicles that cause such social problems. In addition, we present a simple yet effective method to construct an object detection model for the large-scale vehicle images. In particular, we utilize the large-scale of vehicle image sets provided by open AI-Hub, which include the overloaded vehicles from the CCTV, black box, and hand-held camera point of view. We inspected the specific features of sizes of vehicles and types of image sources, and pre-processed these images to train a deep learning-based object detection model. Finally, we demonstrated that the detection performance of the overloaded vehicle was improved by about 23% compared to the one using raw data. From the result, we believe that public big data can be utilized more efficiently and applied to the development of an object detection-based overloaded vehicle detection model.
세선화 기법을 이용한 전역 토폴로지컬 지도의 작성 및 위치 추적
최창혁(Chang-Hyuk Choi),송재복(Jae-Bok Song),정우진(Woojin Chung),김문상(Munsang Kim) 대한기계학회 2003 대한기계학회 춘추학술대회 Vol.2003 No.4
Topological maps have drawn more attention recently because they are compact, provide natural interfaces,<br/> and are applicable to path planning easily. To build a topological map incrementally, Voronoi diagram was<br/> used by many researchers. The Voronoi diagram, however, has difficulty in applying to arbitrarily shaped<br/> objects and needs long computation time. In this paper, we present a new method for global topological map<br/> from the local topological maps incrementally. The local topological maps are created through a thinning<br/> algorithm from a local grid map, which is built based on the sensor information at the current robot position.<br/> A thinning method requires simpler computation than the Voronoi diagram. Localization based on the<br/> topological map is usually difficult, but additional nodes created by the thinning method can improve<br/> localization performance. A series of experiments have been conducted using a two-wheeled mobile robot<br/> equipped with a laser scanner. It is shown that the proposed scheme can create satisfactory topological maps.
이륜차동구동형 로봇의 고속주행을 위한 RRT기반 경로생성기법
문창배(Chang-bae Moon),정우진(Woojin Chung) 대한기계학회 2013 대한기계학회 춘추학술대회 Vol.2013 No.12
In recent days, mobile service robots were widely developed. High-speed navigation is required to improve the task performance. In practical environments, the kinematic and dynamic constraints of a mobile robot should be considered to guarantee the safety for high-speed navigation. In this paper, we propose a motion planning scheme considering the kinematic and dynamic constraints based on the Rapidly exploring Random Trees (RRT). The proposed scheme generates local trajectories using the motion controllers for a two-wheeled mobile robot. By using the identical motion controller, the stable trajectory tracking and safe motion control can be achieved. The simulation results show that the proposed scheme is computationally efficient than that of the conventional RRTs.
동적 장애물의 속도를 고려한 이동로봇의 궤적 재생성 기법
문창배(Chang-bae Moon),정우진(Woojin Chung) 대한기계학회 2014 大韓機械學會論文集A Vol.38 No.11
서비스 로봇이 충돌안전성을 확보한 상태에서 고속 주행 임무를 수행하기 위해서는 동적 장애물의 속도를 고려한 궤적 계획이 필요하다. 정적 장애물만을 고려한 상태에서 궤적을 계획하는 경우 장애물과의 상대속도로 인해서 로봇이 장애물과 충돌할 수 있다. 본 연구에서는 동적 장애물의 속도를 고려한 궤적시간조정기법을 제안한다. 제안된 기법을 통해서 기존에 생성된 궤적의 시간을 조정해서 장애물 회피가 가능한지를 평가할 수 있다. 만일 회피가 불가능할 경우 생성된 경로가 아닌 다른 경로를 선택할 수 있다. 모의 시험 결과를 통해서 제안된 기법을 통해서 짧은 시간 내에 장애물 회피를 수행할 수 있음을 보였다. To achieve safe and high-speed navigation of a mobile service robot, velocity of dynamic obstacles should be considered while planning the trajectory of a mobile robot. Trajectory planning schemes without considering the velocity of the dynamic obstacles may collide due to the relative velocities or dynamic constraints. However, the general planning schemes that considers the dynamic obstacle velocities requires long computational times. This paper proposes a velocity control scheme by scaling the time step of trajectory to deal with dynamic obstacle avoidance problem using the RNLVO (Robot Nonlinear Velocity Obstacles). The RNLVO computes the collision conditions on the basis of the NLVO (Nonlinear Velocity Obstacles). The simulation results show that the proposed scheme can deal with collision state in a short period time. Furthermore, the RNLVO computes the collisions using the trajectory of the robot. As a result, accurate prediction of the moving obstacles trajectory does not required.
딥러닝 영상처리 기반 도로 주행 중 정지 차량 검지 방법
이재용(Jaeyong Lee),정우진(Woojin Jeong),김창일(Chang-il Kim),박진욱(Jinuk Park),박용주(Yongju Park),임용석(Yong-seok Lim) 대한전자공학회 2023 대한전자공학회 학술대회 Vol.2023 No.11
In this paper, we propose a stopped vehicle detection algorithm based on deep learning. Previous studies are focused on detecting stopped vehicle in static situation such as a static camera for road surveillances. In most scenarios for autonomous driving, identifying stopped cars in a driving vehicle is crucial to prevent accidents. However, detecting stopped vehicle needs to estimate the speed of host and observed vehicles which is non-trivial with computer vision. To accomplish this goal, we propose a novel method for speed estimation with road RoI and tracking with deep learning. Also, we utilize optical flow to calculate the movement of the vehicle to which the dashcam mounted. In practice, we implement the program based on the multi-process for efficient computation. The proposed method can effectively recognize stopped vehicles while driving and provide an insight into estimating vehicle speed using in-vehicle dashcams.