A Novel Visual Scale Recovery for Resource-Limited Platform Using Single-Point Laser and Spatial Constraints

Motion estimation of vision-only has no absolute scale, which leads to unmeasurable scale and scale drift. Existing methods of visual scale recovery are either costly or complex, making it difficult to be widely used in resource-limited industries. A lightweight and low-cost method of scale recovery is urgently needed for low-configuration platforms. We innovatively use a camera and an invisible single-point laser transmitter to achieve six-degree-of-freedom (DOF) state estimation, which has minimal sensor requirements for visual scale recovery in terms of size, weight, power, and computational power consumption. It involves the incorporation of a laser into the camera’s field of view and uses a certain spatial relationship between the laser and the camera as a prior constraint for introducing an absolute scale. To address the problem that the laser spot is not visible and cannot be predicted, a background matching method was designed to achieve visual projection matching. To increase the robustness of the engineering, a dual-thread scale estimation method was developed to ensure the continuity and accuracy of scale recovery. The verification conducted with our hardware module and recorded datasets confirmed that the proposed method accurately recovers an absolute scale and controls scale drift. Compared with ORB-SLAM3 and VINS-Mono, which are currently the most concise schemes, the method put forward in this work reduces computational power consumption by more than 54%. Finally, the laser transmitter costs less than U.S.
${\$}$
10. These features render the proposed method an efficient and affordable approach for visual scale recovery.