Sure, I have a pointcloud of a room captured by kinect. Assume that there is a certain object in this pointcloud and I have index of its center. I want to estimate the height of this point with respect to the floor. I also know the location and tilt of the sensor.
Re: Transfering Pointcloud from camera coordinate to world coordinate!
If you know the location and tilt of the sensor in world coordinates, you can then simply build the 4x4 transformation matrix for this pose and then give it to transformPointCloud in order to get your cloud in world coordinates.
But in order to determine where your object is and where the floor is, you will need to add some semantics to your cloud: object recognition, large plane segmentation etc. It is not completely trivial.