Skip to content

Mapping

Antonio Terpin edited this page May 27, 2021 · 1 revision

Mapping parameters

This page summarizes the purpose of the parameters that affect the mapping module. Please, consider the linked references for further details on how the parameters are used by the mapping algorithm.

If you plan to modify or extend the mapping module, consider having a look at the documentation.

Table of contents

  1. Events processing
  2. DSI
  3. Point cloud from voxel grid
  4. Map expansion
  5. Further references

Events processing

Parameter Description
skip_batches amount of batches of events to skip during optimization
skip_batches_for_reset amount of batches of events to skip during optimization after a reset
max_event_rate maximum number of events processed per second; eventually drops events to fulfill this
frame_size number of events to aggregate when computing new map
min_batch_size minimum number of new events required for a map update
events_to_recreate_kf number of events considered when creating a new keyframe to center the pose of the voxel grid

DSI

Parameter Description
min_depth voxel grid minimum depth
max_depth voxel grid maximum depth
num_depth_cells number of depth cells
fov_virtual_camera_deg fov virtual camera (voxel grid)
virtual_width voxel grid width
virtual_height voxel grid height
adaptive_threshold_kernel_size size of the adaptive threshold filter
size adaptive_threshold_c constant offset of the adaptive thresholding

Remark that only the last two parameters are actually private parameters of the module, the others are shared and are described also in the global parameters wiki page

Point cloud from voxel grid

Parameter Description
type_focus_measure type of focus measure to use
radius_search radius filter radius size
min_num_neighbors minimum number of neighbors for radius filter
voxel_filter_leaf_size voxel filter granularity
median_filter_size size of median filter on depthmap
accumulate_local_map_once_every 1 to accumulate all the local maps, n to acccumulate 1 every n
global_point_cloud_skip_first number of local maps to skip when accumulating the global point cloud
radius_search_global_map radius filter on global point cloud radius size
min_num_neighbors_global_map minimum number of neighbors for radius filter on global point cloud

Remark that the global point cloud is not a global map, but it is built accumulating the different local point maps.

The currently available focus measures are summarized in the next table.

type_focus_measure Description
0 Linf: pick 3D points where voxels have a local maximum of votes
1 Contrast: pick 3D points where the contrast of the voxel grid is maximum
2 Gradient magnitude: pick 3D points where the gradient magnitude is maximum

Linf

No extra parameters required.

Contrast

Parameter Description
half_patchsize size of the gaussian filter to estimate contrast

Gradient magnitude

Parameter Description
half_patchsize size of the sobel filter to compute algebraic derivatives

Map expansion

Parameter Description
rate rate at which the node checks whether an expansion is needed
visibility_threshold visibility of the map below which the update is triggered
coverage_threshold minimum amount of pixels covered by reprojected map threshold
baseline_threshold baseline / mean depth ration above which the update is triggered
number_of_initial_maps_to_skip starts checking updates conditions after this number of maps

Further references

To fully understand how the parameters are used in the mapping module, please refer to: