Crate vision_calibration_core 

Core math and geometry primitives for calibration-rs.

This crate provides the foundational building blocks used by all other crates in the workspace:

• linear algebra type aliases (Real, Vec2, Pt3, and friends),
• composable camera models (projection + distortion + sensor + intrinsics),
• a deterministic, model-agnostic RANSAC engine.

Camera pipeline (conceptually): pixel = intrinsics(sensor(distortion(projection(dir))))

The sensor stage supports a Scheimpflug/tilted sensor homography aligned with OpenCV’s computeTiltProjectionMatrix.

Modules

• [math]: basic type aliases and homogeneous helpers.
• [models]: camera model traits and configuration wrappers.
• [ransac]: generic robust estimation helpers.
• [synthetic]: deterministic synthetic data helpers (tests/examples/benchmarks).

Example

use vision_calibration_core::{
    CameraParams, DistortionParams, FxFyCxCySkew, IntrinsicsParams, ProjectionParams,
    SensorParams,
};

let params = CameraParams {
    projection: ProjectionParams::Pinhole,
    distortion: DistortionParams::None,
    sensor: SensorParams::Identity,
    intrinsics: IntrinsicsParams::FxFyCxCySkew {
        params: FxFyCxCySkew {
            fx: 800.0,
            fy: 800.0,
            cx: 640.0,
            cy: 360.0,
            skew: 0.0,
        },
    },
};
let cam = params.build();
let px = cam.project_point_c(&nalgebra::Vector3::new(0.1, 0.2, 1.0));
assert!(px.is_some());

Re-exports

pub use error::Error;

Modules

coordinate_utils

Coordinate transformation utilities for camera projection.

error

Typed error enum for this crate. Typed error enum for vision-calibration-core.

synthetic

Deterministic synthetic data generation helpers.

test_utils

Test utilities for cross-crate calibration testing.

Structs

BrownConrady5

Brown-Conrady 5-parameter radial-tangential distortion model.

Camera

A composable camera model: projection -> distortion -> sensor -> intrinsics.

CameraFixMask

Combined mask for camera calibration parameters.

CameraParams

Serializable camera parameters for building a runtime model.

CorrespondenceView

A single view containing 2D-3D point correspondences.

DistortionFixMask

Mask for fixing distortion parameters during optimization.

FxFyCxCySkew

Standard pinhole intrinsics with optional skew.

HomographySensor

Sensor model represented by a 3x3 homography.

IdentitySensor

Identity sensor model.

IntrinsicsFixMask

Mask for fixing intrinsics parameters during optimization.

NoDistortion

No distortion (identity mapping).

NoMeta

Empty metadata marker for views that do not need extra metadata.

Pinhole

Classic pinhole projection model.

PlanarDataset

A planar dataset consisting of multiple views.

RansacOptions

Configuration parameters for the generic RANSAC engine.

RansacResult

Output of a RANSAC run.

Ray

A camera ray represented by its intersection with the z = 1 plane.

ReprojectionStats

Summary statistics for reprojection errors.

RigDataset

Multi-view dataset for a camera rig.

RigView

One time-synchronized rig frame containing per-camera observations and metadata.

RigViewObs

Multi-camera observations for one rig view/frame.

ScheimpflugParams

Scheimpflug tilt parameters (OpenCV-compatible).

TargetPose

Metadata carrying the per-view pose camera_se3_target.

View

Single-camera observation view with attached metadata.

Enums

DistortionParams

Serializable distortion model parameters.

IntrinsicsParams

Serializable intrinsics parameters.

ProjectionParams

Serializable projection model parameters.

SensorParams

Serializable sensor model parameters.

Traits

DistortionModel

Distortion model mapping between ideal and distorted normalized coordinates.

Estimator

Generic estimator for RANSAC-like methods.

IntrinsicsModel

Intrinsics that map sensor-plane coordinates to pixel coordinates.

ProjectionModel

Projection model from a camera direction to normalized coordinates.

SensorModel

Sensor model mapping between normalized and sensor-plane coordinates.

Functions

compute_mean_reproj_error

Compute mean per-point reprojection error for a calibrated camera and posed views.

compute_rig_reprojection_stats

Compute reprojection error statistics for a multi-camera rig dataset.

compute_rig_reprojection_stats_per_camera

Compute per-camera reprojection error statistics for a multi-camera rig dataset.

distort_to_pixel

Apply distortion and convert to pixel coordinates.

from_homogeneous

Convert a 3D homogeneous vector back to a 2D point.

make_pinhole_camera

Build a PinholeCamera from intrinsics and Brown-Conrady distortion.

normalized_to_pixel

Convert normalized coordinates to pixel coordinates using intrinsics.

pinhole_camera_params

Convert a concrete PinholeCamera into serializable CameraParams.

pixel_to_normalized

Convert pixel coordinates to normalized coordinates using intrinsics.

ransac_fit

Run a generic RANSAC loop for a given Estimator implementation.

to_homogeneous

Convert a 2D point in Euclidean coordinates into homogeneous coordinates.

undistort_pixel

Undistort pixel coordinates to normalized coordinates.

Type Aliases

CameraModel

Concrete camera type built from parameters (f64).

Iso3

3D rigid transform (SE(3)) using Real.

Mat3

3×3 matrix with Real entries.

Mat4

4×4 matrix with Real entries.

PinholeCamera

Concrete pinhole camera alias used across single-camera workflows.

Pt2

2D point with Real coordinates.

Pt3

3D point with Real coordinates.

Real

Scalar type used throughout the library (currently f64).

Vec2

2D vector with Real components.

Vec3

3D vector with Real components.