open3d.t.pipelines.slam.Model

class open3d.t.pipelines.slam.Model

Volumetric model for Dense SLAM.

__init__(*args, **kwargs)

Overloaded function.

  1. __init__(self: open3d.t.pipelines.slam.Model, arg0: open3d.t.pipelines.slam.Model) -> None

    Copy constructor

Parameters:

arg0 (open3d.t.pipelines.slam.Model)

  1. __init__(self: open3d.t.pipelines.slam.Model) -> None

  2. __init__(self: open3d.t.pipelines.slam.Model, voxel_size: typing.SupportsFloat , block_resolution: typing.SupportsInt , transformation: open3d.core.Tensor = [[1 0 0 0], [0 1 0 0], [0 0 1 0], [0 0 0 1]] Tensor[shape={4, 4}, stride={4, 1}, Float64, , device: open3d.core.Device = CUDA:0) -> None

    Constructor of a VoxelBlockGrid

Parameters:

  • voxel_size (SupportsFloat) – The voxel size of the volume in meters.

  • block_resolution (SupportsInt) – Resolution of local dense voxel blocks. By default 16 is used to create 16^3 voxel blocks.

  • transformation (open3d.core.Tensor, optional, default=[[1 0 0 0], [0 1 0 0], [0 0 1 0], [0 0 0 1]] Tensor[shape={4, 4}, stride={4, 1}, Float64) – A 4x4 3D transformation matrix. ()

  • (open3d.core.Device (device) – 0): The CPU or CUDA device used for the object.

  • optional – 0): The CPU or CUDA device used for the object.

  • default=CUDA – 0): The CPU or CUDA device used for the object.

extract_pointcloud(self: open3d.t.pipelines.slam.Model, weight_threshold: SupportsFloat, estimated_number: SupportsInt) → open3d.t.geometry.PointCloud

Extract point cloud from the volumetric model.

Parameters:

  • weight_threshold (SupportsFloat) – Weight threshold to filter outlier voxel blocks.

  • estimated_number (SupportsInt) – Estimated number of surface points. Use -1 if no estimation is available.

Returns:

open3d.t.geometry.PointCloud

extract_trianglemesh(self: open3d.t.pipelines.slam.Model, weight_threshold: SupportsFloat, estimated_number: SupportsInt) → open3d.t.geometry.TriangleMesh

Extract triangle mesh from the volumetric model.

Parameters:

  • weight_threshold (SupportsFloat) – Weight threshold to filter outlier voxel blocks.

  • estimated_number (SupportsInt) – Estimated number of surface points. Use -1 if no estimation is available.

Returns:

open3d.t.geometry.TriangleMesh

get_current_frame_pose(self: open3d.t.pipelines.slam.Model) → open3d.core.Tensor
get_hashmap(self: open3d.t.pipelines.slam.Model) → open3d.core.HashMap

Get the underlying hash map from 3D coordinates to voxel blocks.

integrate(self: open3d.t.pipelines.slam.Model, input_frame: open3d.t.pipelines.slam.Frame, depth_scale: SupportsFloat, depth_max: SupportsFloat, trunc_voxel_multiplier: SupportsFloat) → None

Integrate an input frame to a volume.

Parameters:

  • input_frame (open3d.t.pipelines.slam.Frame) – The frame that contains raw depth and optionally images with the same size from the input.

  • depth_scale (SupportsFloat) – The scale factor to convert raw depth into meters.

  • depth_max (SupportsFloat) – The max clipping depth to filter noisy observations too far.

  • trunc_voxel_multiplier (SupportsFloat) – Truncation distance multiplier in voxel size for signed distance. For instance, –trunc_voxel_multiplier=8 with –voxel_size=0.006(m) creates a truncation distance of 0.048(m).

Returns:

None

synthesize_model_frame(self: open3d.t.pipelines.slam.Model, model_frame: open3d.t.pipelines.slam.Frame, depth_scale: SupportsFloat, depth_min: SupportsFloat, depth_max: SupportsFloat, trunc_voxel_multiplier: SupportsFloat, enable_color: bool = False, weight_threshold: SupportsFloat) → None

Synthesize frame from the volumetric model using ray casting.

Parameters:

  • model_frame (open3d.t.pipelines.slam.Frame) – The frame that contains ray casted depth and optionally color from the volumetric model.

  • depth_scale (SupportsFloat) – The scale factor to convert raw depth into meters.

  • depth_min (SupportsFloat) – The min clipping depth.

  • depth_max (SupportsFloat) – The max clipping depth to filter noisy observations too far.

  • trunc_voxel_multiplier (SupportsFloat) – Truncation distance multiplier in voxel size for signed distance. For instance, –trunc_voxel_multiplier=8 with –voxel_size=0.006(m) creates a truncation distance of 0.048(m).

  • enable_color (bool, optional, default=False)

  • weight_threshold (SupportsFloat) – Weight threshold to filter outlier voxel blocks.

Returns:

None

track_frame_to_model(self: open3d.t.pipelines.slam.Model, input_frame: open3d.t.pipelines.slam.Frame, model_frame: open3d.t.pipelines.slam.Frame, depth_scale: SupportsFloat, depth_max: SupportsFloat, depth_diff: SupportsFloat, method: open3d.t.pipelines.odometry.Method = Method.PointToPlane, criteria: collections.abc.Sequence[open3d.t.pipelines.odometry.OdometryConvergenceCriteria] = [OdometryConvergenceCriteria(max_iteration=6, relative_rmse=1.000000e-06, relative_fitness=1.000000e-06), OdometryConvergenceCriteria(max_iteration=3, relative_rmse=1.000000e-06, relative_fitness=1.000000e-06), OdometryConvergenceCriteria(max_iteration=1, relative_rmse=1.000000e-06, relative_fitness=1.000000e-06)]) → open3d.t.pipelines.odometry.OdometryResult

Track input frame against raycasted frame from model.

Parameters:
Returns:

open3d.t.pipelines.odometry.OdometryResult

update_frame_pose(self: open3d.t.pipelines.slam.Model, arg0: SupportsInt | SupportsIndex, arg1: open3d.core.Tensor) → None
property frame_id

Get the current frame index in a sequence.

property frustum_block_coords

Active block coordinates from prior integration

property transformation_frame_to_world

Get the 4x4 transformation matrix from the current frame to the world frame.

property voxel_grid

Get the maintained VoxelBlockGrid.