该文章提出了一种新的方法，通过因子图优化将平面、属于平面的点和常规3D点联合优化，不依赖深度信息或CNN分割。与RGB-D和Mono Plane SLAM不同，该方法使用最小参数化优化平面，并结合单目平面约束，与ORB SLAM集成。
这篇文章加平面约束，但是没有计算面特征的雅可比，而是使用平面内特征点作为节点参与图优化。
Our formulation, differently from the state of the art, allows us to incorporate general planes, independently of depth information or CNN segmentation being available (although we could also use them)独立于深度信息和CNN
文章主要贡献：
Our main contribution is modelling the joint optimization of planes, points belonging to the planes and regular 3D points using factor graphs.
文章近似的工作有RGB-D Plane SLAM，Mono Plane SLAM,(使用神经网络CNN方法有局限性，限制平面为wall-floor-window)
改写： In our work we address the monocular case, using the minimal parametrization proposed in [12] to optimise planes in the manifold. We combine this with the monocular planar constraint from [16]. Finally, we develop a factor graph-based formulation to optimise planes, points and keyframes and integrate this with ORB SLAM [2].
扫一扫
483
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
