Topic: Best point cloud method w/ Gigapan

Report Abuse Report Abuse
StoneyJr2 (Over 1 year ago)
I have been contemplating the use of photosynth to generate 3D polygon meshes from multiple Gigapans at different angles to the object.  I was thinking of increasing the overlap of the photos to about 50% and choosing angles as you would for photogrammetry.  V-STARS paper can explain.  Is photosynth mostly obtaining the points from the use of the SIFT and the RANSAC algorithms?  Should I include different focal lengths, portrait orientation of sensor along with landscape orientation?  Should I use markers or calibaration bars?  I could measure the edge of a building with my portable laser measuring tape to use for calibration since it would be closer to the size of the subject that I am covering.
StoneyJr2 (Over 1 year ago)
I had tried using focus stacking and Helicon focus to generate a 3D mesh of buildings, but the the specular highlights caused cones to emanate from the mesh where there were direct reflections.
Peter (Over 1 year ago)
You're correct the point cloud is obtained from matched features (thus sparser then point clouds obtained from other sensors like range scanners or dense stereo) To get a decent geometry you will need images from multiple poses , to get parallax, and with overlap. You can use images with differing focal lengths. We don't currently surface a way to incorporate calibration patterns or markers into the geometric reconstruction.