Event

F-LIVO proposes feature-level LiDAR-inertial-visual odometry framework

1 reports · 1 sources

Xinye Dai and other researchers propose F-LIVO, a filter-based LiDAR-inertial-visual odometry framework that couples the visual-inertial and LiDAR-inertial subsystems through feature-level point-to-plane constraints between visual points and LiDAR planes. The source code is planned to be released after the paper is accepted. Experiments on multiple public datasets show that the method's accuracy is competitive but varies across sequences; under the resource configurations evaluated, F-LIVO's average and peak CPU usage, peak resident memory (RSS), and RSS growth rate were the lowest among the evaluated implementations, but R2LIVE's average RSS was lower.