lighting underground / interior advice please

[marcusrjc2]

I am wanting to study lighting requirements for underground photogrammetry for a surveying dissertation / project. I will be using an underground mine so I can eliminate natural light and have full control.

I am hoping to test angles and arrays of lights to test accuracy compared to a laser scan control, using photoscan and cloudcompare software. I have found little research on this and I know diffuse light is advised to eliminate shadows. I would be using 20w LED site (builders) lights with a 120 degree beam.

I am hoping to test using these lights at 90 degress/perpendicular to a wall and then vary this angle with lights either side of a section of wall. I believe varying this angle may vary detail and accuracy of point clouds. I would use multiple lights as I imagine the lighting can't change and so the next secion of wall would have to be lit as I move along and then reposition redundant lights.

I would also be testing the number of light sources for accuracy, e.g. left, right, top, bottom.

Any advice on this would be greatly appreciated as I am relatively new to photogrammetry.

[bigben]

What size tunnel? What angle lens?  Camera pointing along tunnel or perpendicular to walls?  Texture or just geometry?

[marcusrjc2]

The tunnel is about 6m across and 5m high. It's a hard rock (granite) mine. Initially I would just photograph one side wall. I haven't tested a camera yet in this situation to know the lens angle I'd need.

I have tested outdoors and am aware that I need overlap and need to vary camera angles - I've been using about 90 degrees between camera positions on the same area and getting good results.

I'd be shooting approximately perpendicular to the walls but varying camera angles as described above, so really 45 degrees. I'd light an area with 2 lights either side of the area to be photographed and possibly one centrally at a low level. I would need the adjacent sections of the tunnel lit so lighting remains constant on the central section I am photographing. I should be then able to move down the tunnel by moving the first set of lights without affecting the area photographed.

It's the angle of these lights to the wall that I'd like to test for any variation in accuracy of the final point cloud. I am thinking that although you don't want harsh shadows where the dynamic range of the camera won't pick up details, is there an optimal angle?  Does the photogrammetry process need some shadowing to pick up texture or improve how the software picks up corresponding points? I would like to laser scan the same area and compare point clouds using cloudcompare by having ground control points to georeference.

I am a surveying student, hence access to the mine and laser scanner. I have a background in photography, from quite a while ago. As mentioned, photogrammetry is new to me and would be interested to know if this is worth testing and if this methodology is ok.

[marcusrjc2]

I should add that when I refer to area to be photographed, that may be about 5m of tunnel/wall (depending on lens angle) with lights either side and a number of camera angles and overlapping photographs.

[Marcel]

I think the lighting should be as consistent as possible, so either place the lights at a distance and keep them static (if they are strong enough) or set them up with the exact same angle relative to the camera.

Photoscan doesn't need shadow detail specifically, any fine detail will do. Harsh shadows are indeed bad, because often the camera doesn't have enough range to get enough detail in the shadows (or highlights start clipping).

If you photograph a wall you do not need to vary the angles, just varying the position will do. Position variations give Photoscan more information that camera rotation.

Image quality is everything, so use a tripod and a remote release cord, and check your photos afterwards and remove the blurry ones.

[bigben]

I've processed some images shot in long caves http://www.agisoft.com/forum/index.php?topic=3421.msg18794#msg18794  Even with these configurations, a good model of a tunnel would have been possible if they also took photos traveling in the reverse direction.  If the constant banding isn't an issue (i.e. geometry is more important than texture)

Full frame fisheye perpendicular to walls. Lights ahead and behind along axis of tunnel providing light from both sides to reduce shadows. Could be one option.

[marcusrjc2]

Many thanks for thoughts on this. Aprt from practicality of using a fisheye, I'm assuming as it will enable less photographs, are there any other advantages?

Will be using static lights and tripod.

I'm still guessing that minor shadow detail, just to improve texture, may affect results? Guess I will just have to try.

Another thing to test would be the workflow with regard to number of lights. So the more the better if they are static as more even lighting and less moving of lights. Moving down a tunnel I'd light areas 1,2, and 3; photograph area 2; then move area 1 lights to area 4 and photograph area 3 if that makes sense! Hensce just trying a wall to start with. I'd try the whole tunnel later.