학술논문
SLAM(Simultaneous localization and mapping) 기술을 기반 버추얼 프로덕션 카메라 트래킹 방안 연구 / Research on virtual production camera tracking method based on SLAM (Simultaneous localization and mapping) technology
Document Type
Dissertation/ Thesis
Author
Source
Subject
Language
Korean
Abstract
버추얼 프로덕션(Virtual Production)은 창의력과 첨단 기술을 결합한 혁신적인 접근 방식으로, 현대의 영화 및 TV 제작 분야에서 전통적인 영화 및 TV 제작 모델을 점차 재구성하고 있다. 최근 몇 년간 그래픽 처리 장치(Graphics Processing Unit, GPU) 및 기타 주요 컴퓨터 하드웨어의 눈부신 발전과 컴퓨터 생성 이미지(Computer-generated imagery, CGI) 기술의 성숙을 바탕으로, 버추얼 프로덕션 기술은 이제 실시간, 고품질의 캐릭터 및 장면 렌더링을 제공할 수 있는 능력을 갖추게 되었다. 이 기술의 적용으로 효율적인 작업 흐름이 가능해지고, 경제성의 고려와 더불어 창작자에게는 광범위한 예술적 표현 공간을 제공하기 때문에, 이 기술은 널리 주목받고 사용되고 있다. 현재 영화, 드라마, 예능 프로그램, 상업 광고 등 다양한 시각 콘텐츠의 제작과 창작 과정에 버추얼 프로덕션이 폭넓게 자리 잡고 있다. 한편, 동시적 위치추정 및 지도작성 (Simultaneous localization and mapping, SLAM) 기술은, 지난 20년 동안 초기 개념 탐색 단계로부터 고도로 성숙하고 널리 적용되는 기술 프레임워크로 발전해왔다. 이 기술의 연구 및 적용은 초기 학술적 논의에서부터 자율주행 자동차, 무인 항공기, 자율 이동 로봇과 같은 다양한 실제 응용 분야로 점차 확대되었다. 연구가 심화되고 기술이 완성됨에 따라, SLAM 기술은 높은 동적 반응성과 정밀도가 필요한 복잡한 환경에서도 상당한 신뢰성을 가지게 되었다. 다양한 실제 응용 환경에서, SLAM 기술은 이미 그 뛰어난 정확성과 강건함(Robustness)을 입증해왔다.
The virtual production technology has made significant advancements in recent years, particularly during the COVID-19 pandemic, where it has shown great resilience and potential for innovation. With technological progress, virtual production has not only revolutionized visual effects but also provided new possibilities in production processes and methods. One of the core technologies in virtual production is the camera tracking system. However, the camera tracking systems available in the market are often expensive and not widely accessible, limiting the broad application of virtual production technology. SLAM (Simultaneous Localization and Mapping) technology, currently widely used in robotics and autonomous driving, is primarily employed for environmental perception and navigation systems. It constructs environmental maps and locates the device itself by analyzing the motion of the device.This study aims to explore a more cost-effective and widely applicable virtual production camera tracking system using open-source SLAM technology. It focuses on how to apply stereo vision SLAM and stereo IMU multimodal SLAM technologies in the field of virtual production, aiming to reduce the system's cost and complexity while ensuring tracking accuracy and real-time performance. The significance of this research lies in promoting the application of virtual production technology in a broader range of fields, especially beyond the traditionally high-cost and technically demanding areas of film and television production. By integrating open-source technologies and low-cost hardware, we aim to provide a more economical and accessible camera tracking solution for the virtual production field. This will help to stimulate the creative potential of creators while promoting the popularization and development of virtual production technology.Therefore, this study, based on ORB-SLAM3 and VINS-FUSION, two advanced technologies in the open-source SLAM field, proposes a new camera tracking scheme for virtual production. This scheme is not only technically innovative but has also shown high feasibility and efficiency in practical applications through comparative experiments. This research not only provides a new direction for the development of virtual production technology but also offers a practical example of the integration of computer vision and robotics technology, demonstrating the broad prospects of these technologies in artistic creation.
The virtual production technology has made significant advancements in recent years, particularly during the COVID-19 pandemic, where it has shown great resilience and potential for innovation. With technological progress, virtual production has not only revolutionized visual effects but also provided new possibilities in production processes and methods. One of the core technologies in virtual production is the camera tracking system. However, the camera tracking systems available in the market are often expensive and not widely accessible, limiting the broad application of virtual production technology. SLAM (Simultaneous Localization and Mapping) technology, currently widely used in robotics and autonomous driving, is primarily employed for environmental perception and navigation systems. It constructs environmental maps and locates the device itself by analyzing the motion of the device.This study aims to explore a more cost-effective and widely applicable virtual production camera tracking system using open-source SLAM technology. It focuses on how to apply stereo vision SLAM and stereo IMU multimodal SLAM technologies in the field of virtual production, aiming to reduce the system's cost and complexity while ensuring tracking accuracy and real-time performance. The significance of this research lies in promoting the application of virtual production technology in a broader range of fields, especially beyond the traditionally high-cost and technically demanding areas of film and television production. By integrating open-source technologies and low-cost hardware, we aim to provide a more economical and accessible camera tracking solution for the virtual production field. This will help to stimulate the creative potential of creators while promoting the popularization and development of virtual production technology.Therefore, this study, based on ORB-SLAM3 and VINS-FUSION, two advanced technologies in the open-source SLAM field, proposes a new camera tracking scheme for virtual production. This scheme is not only technically innovative but has also shown high feasibility and efficiency in practical applications through comparative experiments. This research not only provides a new direction for the development of virtual production technology but also offers a practical example of the integration of computer vision and robotics technology, demonstrating the broad prospects of these technologies in artistic creation.