feat: add ICameraProjector.TryProjectToPixel (base-frame point → pixel)

[DanielKow]

What

Adds bool TryProjectToPixel(Point3<double> basePoint, out Point<double> pixel) to ICameraProjector / CameraProjector — the inverse of the existing ProjectPoint (pixel → 3D).

public bool TryProjectToPixel(Point3d basePoint, out Pointd pixel)
{
    var cameraToBase = _getCurrentPosition() * _cameraToGripper;
    var pointInCamera = cameraToBase.Inverse().TransformPoint(basePoint);
    if (pointInCamera.Z <= 0) { pixel = default; return false; } // behind/on image plane
    pixel = _intrinsics.PointToPixel(pointInCamera);
    return true;
}

Why

Programs need to draw a known robot-base 3D point (e.g. a teaching point) onto the live camera overlay. The projector already held intrinsics + hand-eye + current pose privately, and CameraIntrinsics.PointToPixel (pinhole + Brown-Conrady) was public, but there was no public base→pixel entry point. This surfaces it without exposing intrinsics.

Returns false when the point is on or behind the image plane (camera-frame Z ≤ 0), where a pinhole projection would otherwise yield a misleading in-front pixel. Callers clip the returned pixel to their actual viewport bounds.

Tests

CameraProjectorTryProjectToPixelTests — 20 strict cases:

• optical-axis point → principal point (exact)
• off-axis points vs. hand-derived pinhole formula (exact, 1e-9)
• on/behind image plane (Z = 0, -0.001, -500) → false + default pixel
• un-project → re-project round trip, no distortion (1e-6) and with distortion (1e-3)
• non-identity gripper pose + non-identity hand-eye vs. independently hand-computed ground truth
• behind-camera under a non-identity pose chain → false
• projection tracks the live gripper pose (camera moves +100mm X → pixel shifts left by exactly Fx·100/Z)

Full RocketWelder.SDK.Automation.Tests: 25/25 green.

🤖 Generated with Claude Code