FusionDepthProcessor Members

The AlignPointClouds function uses an iterative algorithm to align two sets of oriented point clouds and calculate the camera's relative pose. This is a generic function which can be used independently of a Reconstruction Volume with sets of overlapping point clouds.

All images must be the same size and have the same camera parameters.

To find the frame to frame relative transformation between two sets of point clouds in the camera local frame of reference (created by DepthFloatFrameToPointCloud), set the observedToReferenceTransform parameter to the identity.

To calculate the frame-to-model pose transformation between point clouds calculated from new depth frames with DepthFloatFrameToPointCloud and point clouds calculated from an existing Reconstruction volume with CalculatePointCloud (e.g. from the previous frame), pass the CalculatePointCloud image as the reference frame, and the current depth frame point cloud from DepthFloatFrameToPointCloud as the observed frame. Set the observedToReferenceTransform to the previous frames calculated camera pose that was used in the CalculatePointCloud call.

To calculate the pose transformation between new depth frames and an existing Reconstruction volume, pass in previous frames point cloud from RenderReconstruction as the reference frame, and the current frame point cloud (from DepthFloatFrameToPointCloud) as the observed frame. Set the observedToReferenceTransform parameter to the previous frames calculated camera pose.

Note that here the current frame point cloud will be in the camera local frame of reference, whereas the raycast points and normals will be in the global/world coordinate system. By passing the observedToReferenceTransform you make the algorithm aware of the transformation between the two coordinate systems.

The observedToReferenceTransform pose supplied can also take into account information you may have from other sensors or sensing mechanisms to aid the tracking. To do this multiply the relative frame to frame delta transformation from the other sensing system with the previous frame's pose before passing to this function. Note that any delta transform used should be in the same coordinate system as that returned by the DepthFloatFrameToPointCloud calculation.