Processing

Please wait...

Settings

Settings

Goto Application

1. WO2022164165 - DEEP LEARNING TECHNOLOGY-BASED PREDICTION ON POSTURE OF FRONT PEDESTRIAN USING CAMERA IMAGE, AND COLLISION RISK ESTIMATION TECHNOLOGY USING SAME

Publication Number WO/2022/164165
Publication Date 04.08.2022
International Application No. PCT/KR2022/001274
International Filing Date 25.01.2022
IPC
G06N 3/08 2006.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
G06N 3/04 2006.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architecture, e.g. interconnection topology
G06T 7/246 2017.1
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
7Image analysis
20Analysis of motion
246using feature-based methods, e.g. the tracking of corners or segments
CPC
G06N 3/04
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architectures, e.g. interconnection topology
G06N 3/08
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
08Learning methods
G06T 7/246
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
7Image analysis
20Analysis of motion
246using feature-based methods, e.g. the tracking of corners or segments
Applicants
  • 한양대학교 산학협력단 IUCF-HYU (INDUSTRY-UNIVERSITY COOPERATION FOUNDATION HANYANG UNIVERSITY) [KR]/[KR]
Inventors
  • 최준원 CHOI, Jun-Won
  • 김병도 KIM, ByeoungDo
Agents
  • 양성보 YANG, Sungbo
Priority Data
10-2021-001097126.01.2021KR
10-2021-015733016.11.2021KR
Publication Language Korean (ko)
Filing Language Korean (KO)
Designated States
Title
(EN) DEEP LEARNING TECHNOLOGY-BASED PREDICTION ON POSTURE OF FRONT PEDESTRIAN USING CAMERA IMAGE, AND COLLISION RISK ESTIMATION TECHNOLOGY USING SAME
(FR) PRÉDICTION FONDÉE SUR LA TECHNOLOGIE DE L'APPRENTISSAGE PROFOND SUR UNE POSTURE D'UN PIÉTON À L'AVANT À L'AIDE D'UNE IMAGE DE CAMÉRA, ET TECHNOLOGIE D'ESTIMATION DE RISQUE DE COLLISION UTILISANT LADITE PRÉDICTION
(KO) 카메라 영상을 이용한 딥러닝 기술 기반 전방 보행자의 자세 예측 및 이를 활용한 충돌 위험도 추정 기술
Abstract
(EN) Disclosed are deep learning technology-based prediction on a posture of a front pedestrian using a camera image, and a collision risk estimation technology using same. A method for estimating a collision risk by using posture information of a pedestrian, which is predicted on the basis of deep learning technology performed by a posture prediction and collision risk estimation system, according to an embodiment, may comprise the steps of: detecting a pedestrian from each of pieces of image information collected during a predetermined period of time; estimating posture information of the pedestrian during a predetermined period of time, according to a connection relationship between skeletons of the pedestrian for every frame, by using the image information including the detected pedestrian; predicting future posture data and future location data of the pedestrian through a time series analysis of the estimated posture information of the pedestrian during the predetermined period of time; and determining a collision possibility by using the predicted future posture data and future location data of the pedestrian, on the basis of the driving speed and direction of a vehicle.
(FR) Prédiction fondée sur la technologie de l'apprentissage profond sur une posture d'un piéton à l'avant à l'aide d'une image de caméra, et technologie d'estimation de risque de collision utilisant ladite prédiction. Selon un mode de réalisation de l'invention, un procédé d'estimation d'un risque de collision à l'aide d'informations de posture d'un piéton, qui est prédit sur la base de la technologie de l'apprentissage profond exécutée par un système de prédiction de posture et d'estimation de risque de collision, peut comprendre les étapes consistant : à détecter un piéton à partir de chaque élément d'informations d'image recueillies pendant une période prédéterminée ; à estimer des informations de posture du piéton pendant une période prédéterminée, en fonction d'une relation de liaison entre des squelettes du piéton pour chaque trame, à l'aide des informations d'image comprenant le piéton détecté ; à prédire des données de posture future et des données d'emplacement futur du piéton par l'intermédiaire d'une analyse chronologique des informations de posture estimées du piéton pendant la période prédéterminée ; et à déterminer une possibilité de collision à l'aide des données de posture future et des données d'emplacement futur prédites du piéton, sur la base d'une vitesse et d'une direction de conduite d'un véhicule.
(KO) 카메라 영상을 이용한 딥러닝 기술 기반 전방 보행자의 자세 예측 및 이를 활용한 충돌 위험도 추정 기술이 개시된다. 일 실시예에 따른 자세 예측 및 충돌 위험도 추정 시스템에 의해 수행되는 딥러닝 기술을 기반으로 예측된 보행자의 자세 정보를 활용하여 충돌 위험도를 추정하는 방법은, 일정시간 동안 수집된 영상 정보의 각각으로부터 보행자를 검출하는 단계; 상기 검출된 보행자를 포함하는 영상 정보를 이용하여 매 프레임에 대해 보행자의 스켈레톤 사이의 연결 관계에 따라 일정시간 동안의 보행자의 자세 정보를 추정하는 단계; 상기 추정된 일정시간 동안의 보행자의 자세 정보의 시계열 분석을 통해 보행자의 미래 자세 데이터와 미래 위치 데이터를 예측하는 단계; 및 차량의 주행 속도 및 방향을 기반으로 상기 예측된 보행자의 미래 자세 데이터와 미래 위치 데이터에 이용하여 충돌 가능성을 판단하는 단계를 포함할 수 있다.
Latest bibliographic data on file with the International Bureau