Scale Estimation with Dual Quadrics for Monocular Object SLAM
This addresses the scale ambiguity issue for monocular SLAM systems, enabling more accurate 3D mapping in robotics and AR/VR applications, though it is incremental as it builds on existing object-level SLAM methods.
The paper tackles the scale ambiguity problem in monocular SLAM by introducing a scale estimation approach that uses object-level reconstruction and dual quadric representations to align object dimensions with real-world size distributions, achieving absolute scale recovery without prior gravity information.
The scale ambiguity problem is inherently unsolvable to monocular SLAM without the metric baseline between moving cameras. In this paper, we present a novel scale estimation approach based on an object-level SLAM system. To obtain the absolute scale of the reconstructed map, we derive a nonlinear optimization method to make the scaled dimensions of objects conforming to the distribution of their sizes in the physical world, without relying on any prior information of gravity direction. We adopt the dual quadric to represent objects for its ability to fit objects compactly and accurately. In the proposed monocular object-level SLAM system, dual quadrics are fastly initialized based on constraints of 2-D detections and fitted oriented bounding box and are further optimized to provide reliable dimensions for scale estimation.