‘Huis aan de Werf’ testing days, fri. 16 nov.

Post by: Joris Weijdom Add comments
Bodycount

p1010005.jpgOur third day of testing brought us some good results. We decided to start scanning with two camera’s simultaneously, both set up in a slight angle from the middle line (see screenshot). We got some distortion in our data when the projected light line crossed the same height level as the camera was positioned. We think it has something to do with the light projection angle, because the camera is picking up a lot of data slightly above and underneath the camera position.scancabine_topdown_view.jpg

We also get the ‘not detected on the right and left side’ errors when using the ‘2 camera’ setup. This could be an alignment problem. Because the two camera’s are not positioned in an exact 45 degrees towards the scan cabin, the light line is not picked up in level position through the camera lens.

p1010036.jpgIn total, we have scanned with a dozen different software settings today. Major and minor differences resulted in interesting scans. During all scans with a beamer projected light line, we projected in a resolution of 800 x 600 px.
Other minor differences in the soft- and hardware settings during the different scan tests were:

Setup 1.
2 webcams (quickcam pro9000 ) at 1 metre height, more or less default David settings, beamer projecting in a ~ 20 degree angle from the camera, 1 px (reversed) thin shadow line. The cameras are picking up scan data above and underneath camera height level.

Setup 2.
We lowered the cameras to see if we could tackle the angle intersection problem but the whole middle part of the body (torso) was left out on the scan.

Setup 3, 4 and 5.
We performed the scan with another set of clothes, but the dark green pants didn’t showed up on screen. Also, David left out the head and started to collect data from the neck, so we tried a sitting position in which David still left out the head. The third try, He did scan the hands that were resting on the knees quiet nice. It really is a matter of adjusting the settings right.

Setup 6.
We added some contrast and got a slightly better scan

Setup 7.
We tried a 2 px. Line and got an even better scan. A 2pixel light line was about 3 centimetres in height, quiet a thick line, but David did collected data from it.

Setup 8.
p1010009.jpgTo have us all concentrated on the scan procedure, We took some boxes and bags and tried to scan this pile.  The following numbers represent the settings in David from top to bottom: 7027, 2000, 0, 8784, 8468 – 14, 1/20, 1412.
Still, David only picked up a small concentrated part of the setup, the parts that were most frontal to the camera.

p1010025.jpgAfter these eight scans, we discussed the two camera setup and realized we had to get back to a one camera setup, because we kept getting to many errors about intersection angles and alignment. David does get scan information with a 2 camera setup but it would require a separate light line for each individual camera, levelled to the specific camera.

p1010012.jpgWe re-aligned the whole setup and made sure everything was in a level position.
Because so many other scan teams struggle with the right setup and settings (we have seen lots of questions on fora concerning the scan setup), it would be a nice idea for the next patch to add some information regarding the ideal placement of all the objects in settings of different size. Some kind of measurement blueprint.

Setup 9 and 10
Back to a single camera setup. We aligned everything again and made sure the camera was at an exact 45 degrees towards the scan cabin (see top screenshot). Everything has been setup in a straight position. This means with a preview window of 800 x 600 px., we had the middle of the scan cabin at exactly 400 px.

With this last scan setup, we got the best full body scan results of the day. We scanned Arnaud in kaki clothing and got a nice solid mesh out of it.
The settings in David we used (with dim light) were from top to bottom: 6194, 3209, 2164, 8784, 7322 – 0, 1/17, 1255.  (First we keep the exposure setting (1/17) on automatic and see how it will calibrate itself, next we keep these settings and put it on manual again).

p1010049.jpgimg_7957.jpg img_7959.jpgIn the few minutes we had left today, we decided to build a close up setting to see if a smaller scale would have a better outcome. We used two A3 sized papers, we copied the calibration grid on, and taped those to the scan cabin back (at approximately 1,50 metres).

We got some pretty impressive last minute results and were very satisfied with those. A smaller scale scan is much easier to take because you can easily re-align your light line angle by keeping the laser device higher or lower.

p1010059.jpgStill, if the scanning is done manually, this person has to have a very steady hand. It is best if a manual scan is made in one fluid movement from top to bottom or vice versa instead of drawing (moving up and down, like colouring). This way, the laser will pass each facial part only once, and doesn’t show a person’s breathing in the scan result. (see photo)

Because we are confident of the small scale scans, we will use the last test day (next Sunday) to do some full body scale tests again and we will share every new knowledge about it.

I’am aware, the last three posts were quiet technical, but with this knowledge we gained in the past days, we can now answer some of the more compromising questions. How can we give meaning to this created 3d image, can we interact with it, do we feel an emotion or a relation with our image.
Stay tuned for the next post.

Comments are closed.

WP Theme & Icons by N.Design Studio
Entries RSS Comments RSS Log in