<![CDATA[
<p><b><기업정보></b></p><div class="figure-img" data-ke-type="image" data-ke-style="alignCenter" data-ke-mobilestyle="widthOrigin"><img src="https://t1.daumcdn.net/cafeattach/1RgNt/d4f268b0b06c4cd9638509ac3dd73a696fb5ed3b_re_1764052400066" class="txc-image" width="600" height="810" data-img-src="https://t1.daumcdn.net/cafeattach/1RgNt/d4f268b0b06c4cd9638509ac3dd73a696fb5ed3b_re_1764052400066" data-origin-width="814" data-origin-height="1099"></div><div class="figure-img" data-ke-type="image" data-ke-style="alignCenter" data-ke-mobilestyle="widthOrigin"><img src="https://t1.daumcdn.net/cafeattach/1RgNt/7bd828985ee99725038772df3423eec8b9145738_re_1764052400065" class="txc-image" data-img-src="https://t1.daumcdn.net/cafeattach/1RgNt/7bd828985ee99725038772df3423eec8b9145738_re_1764052400065" data-origin-width="649" data-origin-height="1243"></div><p>자율주행 기업에서 SLAM과 멀티모달쪽을 많이 뽑는다.</p><p> </p><p><b><졸업작품 구상></b></p><p>기업쪽에서 SLAM을 많이 뽑으니 ROS 제어와 SLAM 쪽을 해보고싶다.</p><p>SLAM은 LiDAR SLAM과 Visual SLAM이 있는데 우리가 가지고 있는건 2D LiDAR이므로 Dense 카메라를 이용한 Visual SLAM을 해보고싶다.</p><p> </p><p> </p><p><b>(1) slam 이론 공부</b></p><div class="figure-img" data-ke-type="image" data-ke-style="alignLeft" data-ke-mobilestyle="widthOrigin"><img src="https://t1.daumcdn.net/cafeattach/1RgNt/594a2332ba9a3dd3ee799670d0a0ae98a9418ccb_re_1764052400066" class="txc-image" width="300" height="400" data-img-src="https://t1.daumcdn.net/cafeattach/1RgNt/594a2332ba9a3dd3ee799670d0a0ae98a9418ccb_re_1764052400066" data-origin-width="458" data-origin-height="610"></div><p> </p><p><b>(2) SLAM 알고리즘</b></p><p><b><RTAB-Map 알고리즘 (</b><b>R</b>eal-<b>T</b>ime <b>A</b>ppearance-<b>B</b>ased <b>Map</b>ping(실시간 외관-기반 지도생성))></p><p><b>==>카메라로 주변 3D 지도를 자동으로 만들어주는 SLAM 시스템</b></p><p> </p><p><b><a href="https://github.com/introlab/rtabmap" target="_blank" class="ke-link">https://github.com/introlab/rtabmap</a></b></p><div class="figure-open" contenteditable="false" data-ke-type="opengraph" data-ke-align="alignCenter" data-og-type="website" data-og-title="GitHub - introlab/rtabmap: RTAB-Map library and standalone application" data-og-description="RTAB-Map library and standalone application. Contribute to introlab/rtabmap development by creating an account on GitHub." data-og-host="github.com" data-og-source-url="https://github.com/introlab/rtabmap" data-og-url="https://github.com/introlab/rtabmap" data-og-image="https://scrap.kakaocdn.net/dn/SR0gN/hyZNy35Zja/aF5MT1oo4KbuAftoKfELS1/img.png?width=1200&height=600&face=0_0_1200_600,https://scrap.kakaocdn.net/dn/beUfec/hyZOaP0JGm/71TuY1xgaM4swcv3lkck8k/img.png?width=1200&height=600&face=0_0_1200_600"><a href="https://github.com/introlab/rtabmap" target="_blank" data-source-url="https://github.com/introlab/rtabmap"><div class="og-image"><img class="thumb_img" src="https://scrap.kakaocdn.net/dn/SR0gN/hyZNy35Zja/aF5MT1oo4KbuAftoKfELS1/img.png?width=1200&height=600&face=0_0_1200_600,https://scrap.kakaocdn.net/dn/beUfec/hyZOaP0JGm/71TuY1xgaM4swcv3lkck8k/img.png?width=1200&height=600&face=0_0_1200_600" alt="" xxxxonerror="this.src="//img1.kakaocdn.net/thumb/C200x200/?fname=https%3A%2F%2Ft1.daumcdn.net%2Fcafe_image%2Fcafe_meta_image_190529.png""></div><div class="og-text"><p class="og-title">GitHub - introlab/rtabmap: RTAB-Map library and standalone application</p><p class="og-desc">RTAB-Map library and standalone application. Contribute to introlab/rtabmap development by creating an account on GitHub.</p><p class="og-host">github.com</p></div></a></div><p><b><ROS2 환경에서 사용></b></p><p><b><a href="https://github.com/introlab/rtabmap_ros?utm_source=chatgpt.com" target="_blank" class="ke-link">https://github.com/introlab/rtabmap_ros?utm_source=chatgpt.com</a></b></p><div class="figure-open" contenteditable="false" data-ke-type="opengraph" data-ke-align="alignCenter" data-og-type="website" data-og-title="GitHub - introlab/rtabmap_ros: RTAB-Map's ROS package." data-og-description="RTAB-Map's ROS package. Contribute to introlab/rtabmap_ros development by creating an account on GitHub." data-og-host="github.com" data-og-source-url="https://github.com/introlab/rtabmap_ros?utm_source=chatgpt.com" data-og-url="https://github.com/introlab/rtabmap_ros" data-og-image="https://scrap.kakaocdn.net/dn/bDoEkb/hyZOGMXIL5/v8Jj6S8w1kFhUHlYqIL0Vk/img.png?width=1200&height=600&face=0_0_1200_600,https://scrap.kakaocdn.net/dn/b57dUK/hyZOjLkLkc/lZrAfewCFIXP82mt8xEVZK/img.png?width=1200&height=600&face=0_0_1200_600"><a href="https://github.com/introlab/rtabmap_ros?utm_source=chatgpt.com" target="_blank" data-source-url="https://github.com/introlab/rtabmap_ros?utm_source=chatgpt.com"><div class="og-image"><img class="thumb_img" src="https://scrap.kakaocdn.net/dn/bDoEkb/hyZOGMXIL5/v8Jj6S8w1kFhUHlYqIL0Vk/img.png?width=1200&height=600&face=0_0_1200_600,https://scrap.kakaocdn.net/dn/b57dUK/hyZOjLkLkc/lZrAfewCFIXP82mt8xEVZK/img.png?width=1200&height=600&face=0_0_1200_600" alt="" xxxxonerror="this.src="//img1.kakaocdn.net/thumb/C200x200/?fname=https%3A%2F%2Ft1.daumcdn.net%2Fcafe_image%2Fcafe_meta_image_190529.png""></div><div class="og-text"><p class="og-title">GitHub - introlab/rtabmap_ros: RTAB-Map's ROS package.</p><p class="og-desc">RTAB-Map's ROS package. Contribute to introlab/rtabmap_ros development by creating an account on GitHub.</p><p class="og-host">github.com</p></div></a></div><p> </p><p><b><POINT SLAM></b></p><p><a href="https://github.com/eriksandstroem/Point-SLAM?tab=readme-ov-file" target="_blank" class="ke-link">https://github.com/eriksandstroem/Point-SLAM?tab=readme-ov-file</a></p><div class="figure-open" contenteditable="false" data-ke-type="opengraph" data-ke-align="alignCenter" data-og-type="website" data-og-title="GitHub - eriksandstroem/Point-SLAM: Point-SLAM: Dense Neural Point Cloud-based SLAM" data-og-description="Point-SLAM: Dense Neural Point Cloud-based SLAM. Contribute to eriksandstroem/Point-SLAM development by creating an account on GitHub." data-og-host="github.com" data-og-source-url="https://github.com/eriksandstroem/Point-SLAM?tab=readme-ov-file" data-og-url="https://github.com/eriksandstroem/Point-SLAM" data-og-image="https://scrap.kakaocdn.net/dn/zxxKg/hyZOFgdo7U/Sx3YXdfAc1wHkKg7mKbjTk/img.png?width=1200&height=600&face=0_0_1200_600"><a href="https://github.com/eriksandstroem/Point-SLAM?tab=readme-ov-file" target="_blank" data-source-url="https://github.com/eriksandstroem/Point-SLAM?tab=readme-ov-file"><div class="og-image"><img class="thumb_img" src="https://scrap.kakaocdn.net/dn/zxxKg/hyZOFgdo7U/Sx3YXdfAc1wHkKg7mKbjTk/img.png?width=1200&height=600&face=0_0_1200_600" alt="" xxxxonerror="this.src="//img1.kakaocdn.net/thumb/C200x200/?fname=https%3A%2F%2Ft1.daumcdn.net%2Fcafe_image%2Fcafe_meta_image_190529.png""></div><div class="og-text"><p class="og-title">GitHub - eriksandstroem/Point-SLAM: Point-SLAM: Dense Neural Point Cloud-based SLAM</p><p class="og-desc">Point-SLAM: Dense Neural Point Cloud-based SLAM. Contribute to eriksandstroem/Point-SLAM development by creating an account on GitHub.</p><p class="og-host">github.com</p></div></a></div><div class="figure-img" data-ke-type="image" data-ke-style="alignCenter" data-ke-mobilestyle="widthOrigin"><img src="https://t1.daumcdn.net/cafeattach/1RgNt/f2f74ca3f62852e7f3a5e8044b4829e1f36ffb54_re_1764052400066" class="txc-image" width="600" height="391" data-img-src="https://t1.daumcdn.net/cafeattach/1RgNt/f2f74ca3f62852e7f3a5e8044b4829e1f36ffb54_re_1764052400066" data-origin-width="1706" data-origin-height="1112"></div><p>딥러닝 + 3D 포인트 기반 + Dense 카메라로 실행하는 새로운 형태의 SLAM 시스템</p><p>기존 SLAM은 2D특징점 + 조금의 3D 이지만 Point-SLAM은 완전한 3D 뉴럴 point cloud로 map 구성</p><p> </p><p>gpu 연산이 엄청 많아 실시간 불가능</p><p> </p><p><b><mapping 방법 비교></b></p><div class="table-wrap"><table data-ke-type="table" data-ke-align="alignLeft" style="width: 100%; height: 100px;" border="1"><tbody><tr style="height: 18px;"><td style="width: 50%; height: 18px;">RTAB-MAP</td><td style="width: 50%; height: 18px;">Point-SLAM</td></tr><tr style="height: 10px;"><td style="width: 50%; height: 10px;">Surfel 기반 중간-dense map</td><td style="width: 50%; height: 10px;">Neural Feature Point Cloud (학습된 특징 포함)</td></tr><tr style="height: 18px;"><td style="width: 50%; height: 18px;">Occupancy grid</td><td style="width: 50%; height: 18px;">depth noise를 neural fusion으로 제거</td></tr><tr style="height: 18px;"><td style="width: 50%; height: 18px;">Loop closure로 전체 map 보정</td><td style="width: 50%; height: 18px;">Loop closure가 가능하지만 RTAB-MAP 만큼 강하지는 않음</td></tr><tr style="height: 18px;"><td style="width: 50%; height: 18px;">PointCloud는 깊이 기반 그대로 저장</td><td style="width: 50%; height: 18px;">훨씬 smooth하고 정확한 nested point cloud</td></tr><tr style="height: 18px;"><td style="width: 50%; height: 18px;">실시간 처리에 최적화</td><td style="width: 50%; height: 18px;">실시간성은 떨어짐</td></tr></tbody></table></div><p> </p><p>👉 RTAB-MAP = 빠름 / 실시간<br>👉 Point-SLAM = 느리지만 고품질 dense map</p><p> </p><p>따라서 RTAB-MAP으로 SLAM을 할 것임.</p><p> </p><p><b>(3) ROS2</b></p><p>SLAM을 ROS2에 입히기 위해 ROS2 심화를 공부할 예정</p><p>TF2, NAV2 등 다양한 심화를 배워야 SLAM을 ROS2로 구동가능</p><p><a href="https://github.com/orgs/ros2/repositories" target="_blank" class="ke-link">https://github.com/orgs/ros2/repositories</a></p><div class="figure-open" contenteditable="false" data-ke-type="opengraph" data-ke-align="alignCenter" data-og-type="website" data-og-title="ROS 2" data-og-description="Version 2 of the Robot Operating System (ROS) software stack - ROS 2" data-og-host="github.com" data-og-source-url="https://github.com/orgs/ros2/repositories" data-og-url="https://github.com/ros2" data-og-image="https://scrap.kakaocdn.net/dn/cjP3pE/hyZN6tkGHE/qCabMIT4x3pgpep3YlyCh1/img.png?width=280&height=280&face=0_0_280_280,https://scrap.kakaocdn.net/dn/bhORQn/hyZOCKy0YN/zlnRPkUGwxxjo4Cj1gjo91/img.png?width=280&height=280&face=0_0_280_280"><a href="https://github.com/orgs/ros2/repositories" target="_blank" data-source-url="https://github.com/orgs/ros2/repositories"><div class="og-image"><img class="thumb_img" src="https://scrap.kakaocdn.net/dn/cjP3pE/hyZN6tkGHE/qCabMIT4x3pgpep3YlyCh1/img.png?width=280&height=280&face=0_0_280_280,https://scrap.kakaocdn.net/dn/bhORQn/hyZOCKy0YN/zlnRPkUGwxxjo4Cj1gjo91/img.png?width=280&height=280&face=0_0_280_280" alt="" xxxxonerror="this.src="//img1.kakaocdn.net/thumb/C200x200/?fname=https%3A%2F%2Ft1.daumcdn.net%2Fcafe_image%2Fcafe_meta_image_190529.png""></div><div class="og-text"><p class="og-title">ROS 2</p><p class="og-desc">Version 2 of the Robot Operating System (ROS) software stack - ROS 2</p><p class="og-host">github.com</p></div></a></div><p><a href="https://github.com/ros2/geometry2" target="_blank" class="ke-link">https://github.com/ros2/geometry2</a></p><div class="figure-open" contenteditable="false" data-ke-type="opengraph" data-ke-align="alignCenter" data-og-type="website" data-og-title="GitHub - ros2/geometry2: A set of ROS packages for keeping track of coordinate transforms." data-og-description="A set of ROS packages for keeping track of coordinate transforms. - ros2/geometry2" data-og-host="github.com" data-og-source-url="https://github.com/ros2/geometry2" data-og-url="https://github.com/ros2/geometry2" data-og-image="https://scrap.kakaocdn.net/dn/b1UPfE/hyZOGTJfl9/1egNEomC9RNTfpGrvqvMY0/img.png?width=1200&height=600&face=0_0_1200_600,https://scrap.kakaocdn.net/dn/bTjqAd/hyZOkjbLS9/7YbRB9PeoJQvjil2QnFSd0/img.png?width=1200&height=600&face=0_0_1200_600"><a href="https://github.com/ros2/geometry2" target="_blank" data-source-url="https://github.com/ros2/geometry2"><div class="og-image"><img class="thumb_img" src="https://scrap.kakaocdn.net/dn/b1UPfE/hyZOGTJfl9/1egNEomC9RNTfpGrvqvMY0/img.png?width=1200&height=600&face=0_0_1200_600,https://scrap.kakaocdn.net/dn/bTjqAd/hyZOkjbLS9/7YbRB9PeoJQvjil2QnFSd0/img.png?width=1200&height=600&face=0_0_1200_600" alt="" xxxxonerror="this.src="//img1.kakaocdn.net/thumb/C200x200/?fname=https%3A%2F%2Ft1.daumcdn.net%2Fcafe_image%2Fcafe_meta_image_190529.png""></div><div class="og-text"><p class="og-title">GitHub - ros2/geometry2: A set of ROS packages for keeping track of coordinate transforms.</p><p class="og-desc">A set of ROS packages for keeping track of coordinate transforms. - ros2/geometry2</p><p class="og-host">github.com</p></div></a></div><p><a href="https://github.com/ros-navigation/navigation2?tab=readme-ov-file" target="_blank" class="ke-link">https://github.com/ros-navigation/navigation2?tab=readme-ov-file</a></p><div class="figure-open" contenteditable="false" data-ke-type="opengraph" data-ke-align="alignCenter" data-og-type="website" data-og-title="GitHub - ros-navigation/navigation2: ROS 2 Navigation Framework and System" data-og-description="ROS 2 Navigation Framework and System. Contribute to ros-navigation/navigation2 development by creating an account on GitHub." data-og-host="github.com" data-og-source-url="https://github.com/ros-navigation/navigation2?tab=readme-ov-file" data-og-url="https://github.com/ros-navigation/navigation2" data-og-image="https://scrap.kakaocdn.net/dn/coMFH8/hyZN25xlHB/TDYCTupQhXSZKmz7kAs66K/img.png?width=1200&height=600&face=0_0_1200_600,https://scrap.kakaocdn.net/dn/bdQjfX/hyZOvSyOLE/RKGwgPsvKF0mY5pjTapE21/img.png?width=1200&height=600&face=0_0_1200_600"><a href="https://github.com/ros-navigation/navigation2?tab=readme-ov-file" target="_blank" data-source-url="https://github.com/ros-navigation/navigation2?tab=readme-ov-file"><div class="og-image"><img class="thumb_img" src="https://scrap.kakaocdn.net/dn/coMFH8/hyZN25xlHB/TDYCTupQhXSZKmz7kAs66K/img.png?width=1200&height=600&face=0_0_1200_600,https://scrap.kakaocdn.net/dn/bdQjfX/hyZOvSyOLE/RKGwgPsvKF0mY5pjTapE21/img.png?width=1200&height=600&face=0_0_1200_600" alt="" xxxxonerror="this.src="//img1.kakaocdn.net/thumb/C200x200/?fname=https%3A%2F%2Ft1.daumcdn.net%2Fcafe_image%2Fcafe_meta_image_190529.png""></div><div class="og-text"><p class="og-title">GitHub - ros-navigation/navigation2: ROS 2 Navigation Framework and System</p><p class="og-desc">ROS 2 Navigation Framework and System. Contribute to ros-navigation/navigation2 development by creating an account on GitHub.</p><p class="og-host">github.com</p></div></a></div><p> </p><p> </p>
<!-- -->
]]>
카페 게시글
과제게시판
졸업작품 주제 작품 구상 & 공부 계획
한윤지
추천 0
조회 48
25.11.25 16:26
댓글 1
북마크 번역하기 공유하기 기능 더보기
다음검색
첫댓글 1. 멀티모달(영상+포인트클라우드) 객체 검출모델 훈련시 GPU 성능 및 추론시간 조사
2. occupancy network 가 뭔지 조사
3. slam 을 AI로 처리하는게 있는지?
4. 3d lidar로 slam하는 깃허브 찾아볼것 민서랑 차별화 할것 -> 실험은 carla에서 하면됨