Aura and Hovermap onboard processing comparison

Hovermap supports two methods for generating point clouds from mission data: processing in Aura and onboard processing. This article explains the accuracy and speed trade-offs of each method to help you choose the right approach for your project.

Point cloud generation methods

Aura processing runs on a desktop computer after a mission. It uses all lidar rings, performs a global optimization step, and produces a high-resolution point cloud suitable for detailed survey deliverables.

Onboard processing generates a point cloud directly on the Hovermap during or immediately after data collection. It uses a subset of lidar rings, applies more aggressive filtering, and does not perform global optimization. The result is a lower-resolution point cloud (0.1 m voxels) available significantly faster than Aura output.

Processing speed

Onboard processing reduces point cloud generation time by approximately 90% compared to Aura. For a 15-minute mission, Aura requires around 1,200 seconds to process and visualize data. Onboard processing completes the same task in approximately 80 seconds.

Accuracy

Onboard processing produces point clouds that are locally correct but globally less accurate than Aura output. This is because onboard processing does not perform a SLAM global optimization step.

Local vs global accuracy

Local accuracy refers to how correctly points relate to each other over a short distance. For example, a 10-metre section of tunnel wall will still look geometrically correct and undistorted in an onboard processing output.

Global accuracy refers to how well the entire point cloud holds together over the full length of a mission. As the Hovermap moves through a space, small errors accumulate over time. Aura corrects these through a global optimization step that looks back over the entire trajectory and adjusts for accumulated errors across the whole dataset. Onboard processing skips this step to deliver results faster, which means those small errors are not corrected in the same way. Over a long tunnel or complex environment, this difference becomes more noticeable.

In practical terms, onboard processing is well suited to shorter missions and confined environments such as underground stopes. For very long or large-scale missions, Aura is recommended.

Accuracy results

Accuracy was measured by comparing onboard processing output against a reference point cloud generated by Aura. Results are from two underground stope datasets of different sizes and complexity.

For both datasets, the median point cloud error is below 1.1 cm. The higher error at the 95th percentile is mainly due to points measured at longer ranges. The further a point is from the Hovermap, the greater the effect a small position error has on where that point is placed in the final point cloud.

When to use each method

Accuracy limitations to be aware of

Onboard processing accuracy has been validated for underground mining environments. Results in other environments (such as forest or above-ground surveys) have not yet been fully characterized across a range of representative datasets.