Does anyone use photogrammetry apart from those that see the input of many photos, bit of number crunching and out pops a mesh.
I am after an affordable prog that does that but also has the option to mark curves, circles, shapes, lines etc , on aircraft, cars etc. Where the subject lends itself to both engineering type shapes and organic shapes. The mesh types like textured surfaces and bits of aircraft in museums are not always suited being shiny items.
They all seem to be requiring textured objects and masses of photos, no shiny reflections, no smooth surfaces.
What is there other than Photomodeler which does this sort of thing ?
In short: no. photogrammetry will give you, as of now, always a noisy mesh. if you really want smooth surfaces there are two options:
1.smoothing: retopologize (with zbrush, tmap, blender) in a all quad mesh. then use subD or tsplines to smooth. or if you have a good scan (have a look at the agisoft manuals, they are great) try to just smooth it.
2. reverse engineering : rhino reverse, mesh2surface, others
Thanks…yes that sort of thing…
I think and hope I see wireframe output from that program, I am using my own DSLR so will have to contact them as to what all those options are about.
I am on about having to draw on each photo, one starts with reference marks, one marks the same point on two photos, a minimum of 6 required to achieve referenced pics, then with photos referenced one draws lines, curves, etc then references them together, process it and a 3D wireframe results, add more photos to the project, all taken on the same focal length, very laborious, but for subjects which dont have the characteristics suited to the progs that require textures, and overlapping shots, its the only way.
Its what e.g Photomodeler does, and I have taken loads of pics waiting for a time to do this, and need a better prog with tech support.
Manual plotting as such requires photos taken in a different way to the overlapping set a mesh type prog requires, one works around the axis of a curve. Ideally I would like to just take one set of pics and have them suited to manual and mesh type methods.
I’m writing about this at the moment. I found Australis hard to work with, but it looks promising.
The tools that have the most functionality in this area are (surprisingly) tracking software like PFTrack.
Photogrammetry, modeling, being able to place points exactly on corners, supports combining lidar and photogrammetry etc…
I hope Rhino will have better pointcloud support in the future, because it would be the perfect tool for modeling imperfect items from a pointcloud.
I’m struggling to follow the original request a little. Are you looking for a single software that combines both point cloud/mesh scanning and target-based photogrammetry? For the latter approach, have you had a look at Rhinophoto? It requires direct access to the subject as you have to place targets on it; not ideal if for example you have restricted access in an aircraft museum.
What it does do however is allow the user to target (if you can excuse the pun) their focus on the parts of the object that matter to them. The big problem I find with working from mesh scans is that a very high percentage of the data you’re presented with on screen is irrelevant to the task in hand (i.e. making a nurbs model of the subject) and all it does is slow the computer down.
PhotoModeler Scanner works with both discrete points, either applied targets or marked in photos by the user, lines, edges, and circles. It can also be used to create dense point clouds without first identifying discrete points, etc if the surface has sufficient visual texture.
My preference for dense point clouds is PhotoScan.
I import the point clouds and other photogrammetry results into separate .3dm files, and then use those files as references in Worksession for modeling in Rhino.
I would ideally like to find a program maker that has a program with the ability of both cloud/mesh scanning and target based photogrammetry, though I dont ever have the chance to place targets, I have to look for the same features in both photos then reference them together. Museums just dont allow one to go sticking markers on things, and even getting at the subject can be impossible. I might get 15 mins to do a sequence, all the work comes afterwards, it can be a major nightmare where just one marker would be so welcome.
I wonder though if the subject matter would have processed ok, to feed the pics in and have an accurate mesh would be wonderful, but it all needs even lighting, good focus, textured surfaces, situations dont always allow for such.
I am about to trial Photoscan for meshes, and iWitness for the manual method.
I am with Photomodeler but as said, looking for better and with decent tech support.
Trying to find the two methods in one I hope makes the photogrammetry more affordable. I need to find one that accepts one photographic coverage, as to have to take take two sets of photos is terrible, takes long enough photographing round every curve axis there is, 3 shots per axis, to then have to pan around with 60% overlap, hundreds of photos end up getting taken, couple that with flash gun and the museums take a dim view of the disco lighting effect !
to clarify ,
I have two photogrammetry progs, Photomodeler and Agisoft photoscan/now metashape
the first requires a lot of work referencing photos together manually on at least 6 points per photo then requires drawing along curves and straight edges, it doesnt mind reflections and shiny surfaces.
Agisoft requires many more photos, I will wear out my camera with the amount required , and cannot handle shiny surfaces or those with no texture. even lighting without shadows is also more important and not always that easy to achieve especially outdoors in sun or indoors with tall flash.
The first prog you have to have three or more photos around every curves axis, the second requires photos tiling the subject with 60% overlap.
I end up taking photos to suit both.
I am looking for a program that combines both methods saving on not knowing which photo coverage to do. Ideally one that also can take stills from 4K footage as its quicker to film an item than photo it. However one can pull out details in shadow with a raw photo file, not sure about video footage.