http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
번호판 Pose-Estimation을 이용한 주정차 인식 알고리즘 개발
김황근(Hwanggeun Kim),길정환(Jeonghwan Gil),이재석(Jaeseok Lee),최소명(Somyoung Cho),이찬호(Chanho Lee),김철수(Chulsoo Kim) 한국자동차공학회 2023 한국자동차공학회 학술대회 및 전시회 Vol.2023 No.11
Accurately verifying the complete parking status of a vehicle in automated charging and automated equipment is essential. In particular, accurately determining the parking status of a vehicle before initiating tasks such as robot-assisted automatic charging is crucial for ensuring safety. In this study, we propose a parking recognition algorithm using license plate Pose-Estimation. By using the standardized size of the license plate, we accurately measure the distance between the vehicle and the camera using perspective transformation techniques. Based on this, we analyze the movement of the license plate in a specific frame set to quickly determine the parking status of the vehicle. The proposed algorithm is expected to assist in the operation of automated equipment, enhancing overall operational efficiency and safety.
Human-Pose-Estimation을 이용한 Grid 기반 보행자 위험도 인식 기술
김황근(Hwanggeun Kim),길정환(Jeonghwan Gil),이재석(Jaeseok Lee),최소명(Somyoung Cho),이찬호(Chanho Lee),김철수(Chulsoo Kim) 한국자동차공학회 2023 한국자동차공학회 학술대회 및 전시회 Vol.2023 No.11
Pedestrian risk recognition technology is one of the essential tools in automated devices such as autonomous vehicles and collaborative robots to ensure pedestrian safety. Accurately identifying the location of pedestrians using a 2D camera is crucial for this risk recognition. Traditional methods relied on simple techniques that divided the 2D image into Regions of Interest (ROIs) or determined whether an object entered a specific area. Since these methods determine the risk based on the pixel location of the pedestrian in the 2D image without calculating the actual coordinates, errors can occur due to the cameras position, angle, or detection results. In this paper, we introduce a risk recognition algorithm that offers a more accurate approach using Human-Pose-Estimation technology and Auto Calibration. Through Calibration, we correct the cameras distortion and perspective, and using Human-Pose, we calculate the actual coordinates of the pedestrian. By placing the pedestrian in a virtual grid environment, we can evaluate the accessibility to predefined risk areas.