오도메트리 기반 영상 보정 알고리즘을 적용한 동적 영상 매핑에 관한 연구

Abstract

공간 증강 현실(Spatial Augmented Reality)의 한 종류인 프로젝션 매핑은 스마트 폰이나 모니터, TV 등의 2D 디스플레이를 통해 정보를 전달하는 기존의 증강 현실과 달리 건물 외벽, 운동장 바닥, 조형물같이 실제 존재하는 물체를 디스플레이로 활용하여 관찰자에게 극대화된 몰입감을 제공하는 기술이다. 프로젝션 매핑은 움직이지 않는 물체를 대상으로 하는 정적 프로젝션 매핑과 움직이는 물체를 대상으로 하는 동적 프로젝션 매핑으로 나눌 수 있다. 본 논문은 정적 프로젝션 매핑에 비해 다양한 콘텐츠에 적용 가능한 동적 프로젝션 매핑에 대해 연구한다. 기존의 동적 프로젝션 매핑은 대상 물체와 프로젝터의 기하관계를 찾기 위해 카메라와 같은 비전센서를 사용한 연구가 주를 이루지만 실제 물체와 가상의 물체가 정확히 매핑 되지 않을 경우, 완성도 높은 콘텐츠를 요구하는 상업용 공연, 전시 분야에 적용하기 곤란하며 이를 해결하기 위해서는 고가의 물체 인식 장비를 사용해야 하는 문제점이 있다. 본 논문에서는 상업용 콘텐츠에 적용하기 위해 고정밀 모터나 로봇-암의 오도메트리 정보를 이용한 동적 프로젝션 매핑 기법을 제안한다. 오도메트리 기반 동적 프로젝션 매핑은 카메라 같은 비전센서를 사용한 동적 프로젝션 매핑에 비해 물체의 움직임이 제한적이지만 고가의 물체 인식 장비를 사용하지 않고도 완성도 높은 동적 프로젝션 매핑의 연출이 가능하기 때문에 다양한 상업용 콘텐츠에서의 적용이 가능하다. 제안하는 오도메트리 기반의 동적 프로젝션 매핑에서의 프로젝션 매핑 대상은 고정밀 모터나 로봇-암 끝단에 설치되는데, 설계 후 가공되거나 3D프린터로 제작한 기하 정보를 알고 있는 물체가 사용된다. 이러한 물체의 기하 정보와 동적으로 변하는 오도메트리 정보, Light Sensor를 사용하여 프로젝션 기반의 캘리브레이션을 수행한다. 또한, 오도메트리 기반의 캘리브레이션 결과, 프로젝터, 대상 물체의 기하학적 관계를 통해 물체와 모터 축의 조립 시 발생하는 회전축의 오차 기울기를 찾고 보정함으로써 시각적 오차를 줄인다.| Projection mapping, a type of Spatial Augmented Reality, can provide maximized immersion to an observer by utilizing an actually existing object, such as an exterior wall of a building, field ground, and 3D sculpture through the display and is unlike existing augmented reality wherein information is delivered via a 2D display, such as a smartphone, monitor, or TV. The projection mapping can be classified into static projection mapping, which focuses on a still object, and dynamic projection mapping, which focuses on a traveling object. The present study regards dynamic projection mapping, which can be applied to various contents compared to static projection mapping. A majority of existing studies on dynamic projection mapping focus on vision sensors including camera, to find the geometric relationship between the object and projector. However, when the actual object and virtual object are not accurately mapped, their application to professional fields such as commercial shows or exhibitions is difficult, and this leads to the use of costly objection-recognition equipment. Suggested in the present study is a static projection mapping that uses the odometry information of a high-precision motor or robot arm for application to commercial contents. Although odometry-based dynamic projection mapping is limited with the object’s movement as compared to dynamic projection mapping, which utilizes vision sensors like camera, because odometry-based dynamic projection mapping is capable of materializing a top-notch dynamic projection mapping, this possesses a quality that can even be applied to various commercial contents. The target of the suggested odometry-based dynamic projection mapping is installed on the edge of either a high-precision motor or robot arm. The target’s geometric information is provided to the user by utilizing its design information or a 3D printer. By utilizing the target object’s geometric information and dynamic odometry information and light sensor, a projection-based calibration is performed. Moreover, the relationship between the result of the odometry-based calibration, projector, and geometric information of target object, the error gradient of rotation axis, which occurs during the assembly of object and motor axis, can be calibrated, thus reducing visual errors.; Projection mapping, a type of Spatial Augmented Reality, can provide maximized immersion to an observer by utilizing an actually existing object, such as an exterior wall of a building, field ground, and 3D sculpture through the display and is unlike existing augmented reality wherein information is delivered via a 2D display, such as a smartphone, monitor, or TV. The projection mapping can be classified into static projection mapping, which focuses on a still object, and dynamic projection mapping, which focuses on a traveling object. The present study regards dynamic projection mapping, which can be applied to various contents compared to static projection mapping. A majority of existing studies on dynamic projection mapping focus on vision sensors including camera, to find the geometric relationship between the object and projector. However, when the actual object and virtual object are not accurately mapped, their application to professional fields such as commercial shows or exhibitions is difficult, and this leads to the use of costly objection-recognition equipment. Suggested in the present study is a static projection mapping that uses the odometry information of a high-precision motor or robot arm for application to commercial contents. Although odometry-based dynamic projection mapping is limited with the object’s movement as compared to dynamic projection mapping, which utilizes vision sensors like camera, because odometry-based dynamic projection mapping is capable of materializing a top-notch dynamic projection mapping, this possesses a quality that can even be applied to various commercial contents. The target of the suggested odometry-based dynamic projection mapping is installed on the edge of either a high-precision motor or robot arm. The target’s geometric information is provided to the user by utilizing its design information or a 3D printer. By utilizing the target object’s geometric information and dynamic odometry information and light sensor, a projection-based calibration is performed. Moreover, the relationship between the result of the odometry-based calibration, projector, and geometric information of target object, the error gradient of rotation axis, which occurs during the assembly of object and motor axis, can be calibrated, thus reducing visual errors.