MakerBot Industries

I am having a very similar problem. As far as I can tell I am doing everything correctly. When I open the point cloud in meshlab I see a flat plane, the blank wall behind the object, and I see the shadows of the object, (camera view angle shadow, and laser shadow) and sometimes I see a 2d representation of the object. Most times it's just random points.

Please help me figure out what I'm doing wrong.

Screen shots and the point cloud can be see here:

tinyurl dot com slash 72ewqup

I'm using the PS3 camera from the wiki, the printed parts, and a 5mW laser. I have tried different light sensitivity settings, different backgrounds, and different floors.

Hi rlspell,

From your screenshot of the scanning process I can see that something's not right. The red part of the line (which denotes where the laser is being detected at the current moment) should be essentially vertical except for jumps where the 3D geometry changes. In your screenshot, you can see that there are red pixels on the edges of the object and (worse) on the left edge of the screen. These red pixels should be where the laser is appearing instead. This could be caused by a few things:

1. Your camera moved between the first static picture (taken right at the beginning of the scan) and the current time.
2. The object surface is bright causing reflections on its edges
3. The lighting changed between the static image the current frame
4. The static picture was taken with the laser on during the "cover the laser" portion of the scan

All of these have to do with the way the laser is detected. I use an image-difference technique between an image taken at the beginning (when you covered the laser, ie: http://makerbot.wdfiles.com/local--files/makerscanner-software/laserCovered.png) and the current frame. Thus, the idea is that the only thing that has changed is where the laser is, so I can quickly and accurately compute the laser's position.

In your case, the image difference algorithm is showing a large difference on the left edge, so somehow that part of the image has changed dramatically (more so than the part of the image where the laser is).

Let me know if all of that makes sense and if you're still stuck. Also, the screenshot is very helpful, so definitely keep posting them if you're still having trouble.

Best,

Andy

Thanks for the quick reply.

The red pixels in the screen shot showing scanning (with all the blue lines) *IS* the laser. Maybe it's not bright enough? It's a 5mW laser, and I did turn the thresholds down all the way and try it there.

Laser is within a couple of degrees of vertical, as you can see from the first "laser" screenshot. (original link above should still work)

1) Camera mount (and laser mount, plastic parts) are all hot glued to 3/4" board, they are not moving.
2) I tried some reflective items, and some black items, same results. this particular item is a dull white plastic.
3) No lighting changes, just repeated with the same object, further away. Also just tried it butted up against the back wall.
4) I don't cover the laser, turn it on after the "cover the laser" portion. still, the static picture doesn't change.

The large number of blue pixels on the left of the original scan is where the laser is reflecting off the back wall. they are on the left because the laser is on the right. In that particular scan I ran the laser back and forth a bunch of times to get lots of blue pixels on the scanned object, so of course the wall directly in line with the laser and object got lots of pixels.

Object was halfway between wall and camera. I have tried different positions all with the same results.

Thoughts?

using your description, that it works on a differential image theory, I grabbed the cord of the item I'm scanning and pulled it to the right 1/2" during scanning. You would think it would then be completely covered in blue, both where it was, and the new position. Not so. See screenshot at original link.

Hi rlspell,

Sorry, I should have been more clear. I was talking about the pixels on the very left, (a single vertical column) that are red. Perhaps there is a bit of noise introduced by the camera driver there, since the algorithm keeps identifying that as having a large change. Does the very left column of pixels flicker at all when you just view the camera?

Screenshot-laser2.png looks right — the red line is where it should be, except for at the bottom of the image when it's jumped back to the left. The red line in that view isn't actually showing the laser, it's showing where the detection algorithm has decided the laser is, and thus where the 3D geometry is being calculated form. The image behind it is also static — it's the one captured when the laser is off. Once that looks right, the pointcloud will be correct because they are directly related.

So basically, look to see if there could be an issue on the camera on the left side and then if so we'll come up with some way to either ignore those pixels or fix the camera driver.

Sound good?

- Andy

P.S. regarding the differential image theory, I'm 100% sure it works that way because I wrote the code for it :-)

ok, I see that now. I'll fiddle with different cameras and maybe get a new PS3 one to try. I'll have to dig around and see if the Ubuntu driver needs updating. I'll also try different USB cables, ports, etc. Thanks for the help, and I'll post an update when I figure it out. :-)

Ok, sounds good. The other option is to change the code to ignore that column of pixels, which shouldn't be too bad. Let me know.

- Andy

I have tried six different cameras (all the ones recognized by Ubuntu) and three different usb ports to no avail. There appears to be no flickering just looking at the image. Unless I turn the lights off, then there is noise from the cameras. I have even gone so far as to mount a 60 watt light over the image to be scanned, in the hopes of increasing the signal to noise ration. No effect, still the red line to the left.

If I had wrote it and I did a line by line scan of the image I would suspect a bug, as it's the first part of each line, but I'm sure your code is fine, and I have no idea how you are comparing the images.

How do you capture the static image? Just grab one frame? Maybe during the "cover the laser" portion, grab bunches of frames and compare them, and filter out the pixes that change as "noisy pixels"…? You have probably already thought of that.

I will continue to try to clean up the image, but the static picture just looking at it is pretty clear with no flickering that I can see on the left border.

What would it take to filter out the noisy column on the left?

was able to eliminate the red column on the left, but turning brightness threshold down all the way, and brightness filter up all the way.

scanned object.

point cloud still problematic.

screen shots in the original link

thoughts?

also had to move the paper over so it wasn't on the far left of the image at all…

Hi rlspell,

Ok, I looked at the code and there might have been an issue in how I was handing the "no data" case (ie places where the laser line did not appear) that could have caused it to produce the issues you're seeing. I don't have a scanner here to test with, so if you want to give the fix a try, that'd be great. Just download this new version of ScanThread.cpp here: http://makerscanner.com/files/ScanThread.cpp

Replace your old ScanThread.cpp with it, then recompile.

Let me know either way, and sorry for the delay.

Best,

Andy

Comment removed.

Thanks very much. I'll post again when I have some results.