Point Cloud Bounding Box to 2D Bounding Box

Description

This operator is used to project a 3D point cloud bounding box onto a 2D image pixel bounding box via the camera intrinsic matrix, and synchronously generate the corresponding 2D ROI.

Overview

Projects 3D boxes into 2D boxes in the selected camera.

Scenarios

Commonly used in scenarios such as the fusion of vision and point cloud results, and visualization of detection results, to annotate the region corresponding to the point cloud bounding box on the image.

Inputs and Outputs

Inputs

Point Cloud Bounding Box: 3D bounding box in the camera coordinate system.

Image (Optional): Optional RGB image used for visualization output.

Camera parameters: Camera intrinsics, distortion, color-camera, and passive-stereo parameters.

Outputs

Pixel bounding box: Four-point pixel box corresponding to each point cloud bounding box, formatted as N×4×2.

2D ROI: Rectangular region corresponding to each bounding box.

Parameters

Image Source

Description

Selects which camera calibration is used for projection or reconstruction. Auto infers it from the input image where supported.

Tuning Instructions

Use Auto when an image is available and identification is reliable; otherwise choose the camera to which the pixels, depth, or desired projection belong.

Parameter Range

Options: Auto, Texture camera, depth camera, Default: Auto

Parameter Range

Values/Range:

  • Auto

  • Texture Camera

  • Depth Camera

Default: Auto