Modern mobile robotic systems depend on sensor fusion for safety and redundancy. While the object detection, planning, and mapping algorithms that depend on sensor fusion get all the attention, it is often forgotten that their success depends on accurate sensor calibration.
Spleenlab VISIONAIRY® Online Calibration provides system designers an out-of-the-box solution to solve the challenging and overlooked calibration problem and reduce the maintenance requirements and cost for complex sensor stacks.
Recent advances in deep neural network-based monocular camera depth estimation have opened the possibility of treating cameras as pseudo-depth sensors in a variety of applications. Spleenlab’s online calibration algorithm uses this technology to elegantly calibrate cameras to LiDAR sensors without the need for targets or operator interaction, eliminating the need for expensive and time-consuming target-based calibrations in the workshop.
One of the most challenging calibration problems to solve is calibrating the camera to the vehicle itself (ex. the vehicle’s rear axle). This calibration is critical for fusing information from the camera into the vehicle-centric sensor fusion framework found in robotic systems.
A problem that has already been solved in large part by LiDAR manufacturers, is the calibration of the LiDAR to the vehicle. Using Spleenlab’s online calibration algorithm, each camera can be calibrated to the LiDAR sensor first, and then by association to the vehicle. In essence using the LiDAR sensor and surrounding environment as the calibration target itself, allowing for all cameras to be calibrated to the vehicle using a simple automatic target-free calibration algorithm that works in all environments.
Spleenlab VISIONAIRY® Online Calibration is a leader in the continuous online extrinsic/intrinsic calibration (COEC/COIC) domain required for the successful deployment of robotic vehicles at scale; minimizing platform downtime, building a safer system, and guaranteeing a VISIONAIRY future.
