🟠 Point Cloud to Mask Image

Function Description

The operator receives one or more point clouds and uses camera intrinsic matrix to project these 3D points onto the 2D image plane, generating a corresponding binary mask image for each input point cloud. The final generated mask is a filled solid region containing all projected points.

Usage Scenarios

By projecting calculated 3D point clouds back to 2D images, you can find the corresponding region of the 3D object on the 2D image, intuitively checking whether the projected contour fits the actual object edges in the image.

Input Output

Input

Image: A reference image used to define the dimensions of the output mask.

Point cloud: A list of point clouds, each will be converted to an independent mask image, generally input unordered point clouds after ROI or segmentation.

Camera intrinsics: 3x3 camera intrinsic matrix used to project 3D points to 2D pixel coordinates.

Output

Mask image list: A list of binary mask images with the same count as the input point cloud list elements.

Parameter Description

Coordinate system requirements: The input "point cloud" must be in the camera coordinate system. Since the operator uses camera intrinsics for projection, incorrect coordinate systems will lead to inaccurate projection results.

Take Minimum Bounding Box

Parameter Description

Controls whether the final generated mask region tightly fits the contour or is a regular rectangle.

Parameter Adjustment

  • Disable (default): The generated mask is a polygon that tightly wraps all projected points, with shapes more fitting object contours and higher precision.

  • Enable: The generated mask is a minimum area rectangle that can encompass all projected points.