Struct PointCloud3D 

pub struct PointCloud3D {
    pub points: Vec<[f32; 3]>,
    pub normals: Option<Vec<[f32; 3]>>,
    pub colors: Option<Vec<[u8; 3]>>,
    pub intensities: Option<Vec<f32>>,
}

A 3-D point cloud.

Points are stored as [f32; 3] (x, y, z). Optional per-point attributes (normals, colours, intensities) are stored as parallel Vecs.

Fields

points: Vec<[f32; 3]>

[x, y, z] positions.

normals: Option<Vec<[f32; 3]>>

Unit surface normals — populated by Self::estimate_normals.

colors: Option<Vec<[u8; 3]>>

Per-point RGB colour in [0, 255].

intensities: Option<Vec<f32>>

Per-point scalar intensity.

Implementations

impl PointCloud3D

pub fn new(points: Vec<[f32; 3]>) -> Self

Create a new cloud from a Vec of [x, y, z] points.

pub fn from_depth_map(depth: &[Vec<f32>], intrinsics: &CameraIntrinsics) -> Self

Unproject a depth map into a point cloud using pinhole intrinsics.

Pixels with depth ≤ 0 are skipped. The resulting cloud is in the camera frame.

Arguments

• depth – [row][col] depth map in metres.
• intrinsics – Camera intrinsics for unprojection.

pub fn len(&self) -> usize

Number of points in the cloud.

pub fn is_empty(&self) -> bool

Returns true when the cloud contains no points.

pub fn centroid(&self) -> [f32; 3]

Centroid (mean position) of the cloud.

pub fn bounding_box(&self) -> ([f32; 3], [f32; 3])

Axis-aligned bounding box: (min_xyz, max_xyz).

pub fn remove_statistical_outliers(&mut self, k: usize, std_ratio: f64)

Remove statistical outliers (SOR filter).

Each point’s mean distance to its k nearest neighbours is computed. Points whose mean distance exceeds mean + std_ratio * std_dev of the global distribution are removed.

self is mutated in place. All optional attribute arrays (normals, colors, intensities) are filtered to match.

pub fn voxel_downsample(&self, voxel_size: f32) -> Self

Downsample the cloud by replacing all points inside each voxel cell with their centroid.

Returns a new cloud. Optional attributes are NOT preserved (they would require averaging, which depends on the attribute semantics).

pub fn estimate_normals(&mut self, k: usize)

Estimate per-point surface normals by PCA on the k-nearest neighbour neighbourhood.

The smallest eigenvector of the 3×3 covariance matrix of the neighbourhood is taken as the normal. Orientation is made consistent by flipping to point “upward” (positive Z component).

self.normals is populated (or replaced) in place.

pub fn icp_align( source: &PointCloud3D, target: &PointCloud3D, max_iterations: usize, tolerance: f64, ) -> (PointCloud3D, [[f64; 4]; 4])

Align source to target using point-to-point ICP.

Returns the aligned source cloud and the 4×4 cumulative transform.

The algorithm:

1. For each source point find the nearest target point.
2. Compute the optimal rigid transform via SVD of the cross-covariance.
3. Apply the transform and update the cumulative matrix.
4. Stop when the mean correspondence distance change is < tolerance or max_iterations is reached.

pub fn k_nearest_neighbors(&self, query: [f32; 3], k: usize) -> Vec<usize>

Return the indices of the k nearest points (excluding self).

pub fn to_ply_string(&self) -> String

Serialise the cloud to an ASCII PLY string.

pub fn from_ply_str(s: &str) -> Result<Self, String>

Parse an ASCII PLY string into a PointCloud3D.

Supports property float x/y/z, property float nx/ny/nz, and property uchar red/green/blue.

Trait Implementations

impl Clone for PointCloud3D

fn clone(&self) -> PointCloud3D

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for PointCloud3D

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.