- Thread starter Smouser
- Start date
Screwdriver
Member
- Messages
- 10,710
Measure it. Scanning won’t give you anything like the accuracy you’re after.
hotponyshoes
Member
- Messages
- 7,450
- Location
- Somerset. Uk
Something I've been involved with recently is 'scanning' horses feet. You then 3d print a shoe and glue it on.
Firstly, the layer of glue is 3-4mm thick so that's the level of accuracy required for this
I'm not involved with the software side so I have no idea how it works but what we have been doing is using a reference marker for a known size scale.
This is basically a credit card with a circle and a cross cut out.
You put the card on the floor by the foot.
Rather than a scanner you take multiple photos (or a video) moving around the foot and keeping the card in view.
It's then processed by the software and it works out the angle of the view based on the distortion of the shapes on the card.
It's AI based and there is not much manual input.
The model produced includes a model of the card so its easy to check that the hole in the card is still 20mm for example.
I don't know how this would relate to scanning precision objects but maybe you could do something like you would do on a cmm table.
Mount the object you want to scan.
Mount a 20mm (whatever) diameter sphere/ball bearing at each corner.
Scan the lot.
Check your generated model includes the bearings and each are 20mm in all directions.
If not re-scale as appropriate.
I'm only scanning about 220deg not going the full 360 so I've only been using one marker card but I don't see why multiples wouldn't work.
For scanning objects that you could put on a turntable you could put scale markings on the table itself.
Or,
Measure the table accurately.
Place some known-size objects on it.
Scan the lot.
Create some software that calibrates the scan against the known measurements and adjusts ongoing scans against that.
I'm not sure what factors limit the precision of 3d scanning but, if a farrier with no skills and a mobile phone can get +/- 1mm on a 150mm+ object, you clever lot should be able to work out a cheap & easy way to get 0.1mm at least...
Firstly, the layer of glue is 3-4mm thick so that's the level of accuracy required for this
I'm not involved with the software side so I have no idea how it works but what we have been doing is using a reference marker for a known size scale.
This is basically a credit card with a circle and a cross cut out.
You put the card on the floor by the foot.
Rather than a scanner you take multiple photos (or a video) moving around the foot and keeping the card in view.
It's then processed by the software and it works out the angle of the view based on the distortion of the shapes on the card.
It's AI based and there is not much manual input.
The model produced includes a model of the card so its easy to check that the hole in the card is still 20mm for example.
I don't know how this would relate to scanning precision objects but maybe you could do something like you would do on a cmm table.
Mount the object you want to scan.
Mount a 20mm (whatever) diameter sphere/ball bearing at each corner.
Scan the lot.
Check your generated model includes the bearings and each are 20mm in all directions.
If not re-scale as appropriate.
I'm only scanning about 220deg not going the full 360 so I've only been using one marker card but I don't see why multiples wouldn't work.
For scanning objects that you could put on a turntable you could put scale markings on the table itself.
Or,
Measure the table accurately.
Place some known-size objects on it.
Scan the lot.
Create some software that calibrates the scan against the known measurements and adjusts ongoing scans against that.
I'm not sure what factors limit the precision of 3d scanning but, if a farrier with no skills and a mobile phone can get +/- 1mm on a 150mm+ object, you clever lot should be able to work out a cheap & easy way to get 0.1mm at least...
I have no illusions nor am I a fanboy for phone/webcam photogometry, as some people will know already but...
I have a raptor cr with the scanbridge so it can be used wirelessly & in blue laser mode backed off against some other stuff when I checked it, its within 0.2-0.25mm and you dont have to do anything special to get that apart from apply marker dots to the piece for tracking. The manufacturer (creality) claims 0.1mm, but thats a "ideal" figure. They offer 2 different models called the raptor-x and raptor-pro that claim 0.02mm resolution, and these have crossed blue lasers in place of the straight lines of mine, and they are supposed to be better for hair/materials and difficult objects (ie shiny metal). You can coat parts in a scanning spray that subliminates (evaporates leaving no residue in x minutes) to work around this, though I hear rumours that dry hair shampoo works too that I will test because scanning spray is 35 quid a can for the aesub blue.
So Its not metrology grade despite their claims and the software is a bit buggy and it needs a really fast high end pc, but within 0.2mm is enough for the vast majority of work, building fixtures to hold things, car trim, stuff you will 3d print anyway because well, its plastic squirted out of a computer controlled plastic glue gun. Maybe as I get more seat time on it, I can get nearer that 0.1mm as it definitely is a skill you have to aquire.
Your not going to use it to align something to micron accuracy, but it is what it is.
There are some programs that can mesh the point cloud and extract geometry for a traditional vector drawing of a part, but still investigating and learning that as a vector cad program is more use than a point cloud in a lot of circumstances.
Posted elsewhere here already, biggest scan Ive done so far, a rover v8 4.6L block. The orange areas represent insufficient data, but that was because the scan spray subliminated off the liners/inner crank web area before I got to them, and if you re-apply it blinds all the marker dots making tracking & scanning impossible. Its perfectly adequate to make some fixtures and for designing engine mounts, exhausts etc in future from.
I've scanned and 3dp some smaller things direct from the point cloud model exporting to stl then slicing it already, a replacement hydromassage jet for the wife's spa and the funny little twist to lock fixtures that hold the end fuse panel cover on the slk dash as some were missing.
You can scan in just point data (the rover block), or add a scan taken from the visual image (like in the slk console scan which shows the marker dots up visually), the rover I didnt need visual clues and it cuts down on processing size, its already at nearly a million points in the cloud with just raw data without image mapping.
Post processing I spend some time manually removing stray dots/noise which all photogommetry suffers from to some degree (including the 100k industrial solution my son uses in his job as a industrial metrology technician...), then rubbish removed, I proces it into a point cloud and mesh it, then if there's some areas not well defined or to scan from underneath etc (the rover v8 is a true 360 model, you can rotate it round in 3d in cad and see into it from the sump etc) take more scans and add them the project. Then I pick markers on the two scans that are in two or three planes & show up in both scans and use these to align the scans. So I put markers & I have some tall marker targets on magnets to "walk" around corners to get a extra plane of data for this. Feature alignment or non marker mode is a joke, it always messes up, but manually guided by a mk1 eyeball its really accurate and works well. Once you have them aligned the software asks if its ok, and you merge it into one point cloud model. Then you can take more scans, and repeat adding them the model each time, or in one giant mergathon at the end if you prefer.
I think the rover block in the end was 10-12 separate scans merged into one model but that was a complex awkward thing to scan, normally I take 2-3 scans of something to scan it in 3d. If I needed the liner detail I could always go back and mask off the tracking dots and spray it with scan spray and repeat to eliminate the orange areas, but its not needed.
The slk plastics were a nightmare to scan at first, but once the "soft touch" plastic spray coating put on at the factory had been removed (which was needed anyway, it had turned to sticky snot with age like a lot of merc/bmw interiors of the era), they didnt need spray to work apart from the odd bit I missed a bit of soft touch coating on somewhere hidden where a quick sniff made life easier.
I spent about 3 weeks just 3dprinting marker towers with my resin printer and sticking target dots on them and gluing neo magnets into them, going cross eyed and used 2500 marker dots but a worthwhile investment in time I think, and they are suprisingly expensive to buy. You can see some of these stuck the shelving in the background in the actual photo of the rover block on the bench.
I havent used it in IR mode, that seems a bit playtown toy for scanning faces with no reference geometry to me.
I have a raptor cr with the scanbridge so it can be used wirelessly & in blue laser mode backed off against some other stuff when I checked it, its within 0.2-0.25mm and you dont have to do anything special to get that apart from apply marker dots to the piece for tracking. The manufacturer (creality) claims 0.1mm, but thats a "ideal" figure. They offer 2 different models called the raptor-x and raptor-pro that claim 0.02mm resolution, and these have crossed blue lasers in place of the straight lines of mine, and they are supposed to be better for hair/materials and difficult objects (ie shiny metal). You can coat parts in a scanning spray that subliminates (evaporates leaving no residue in x minutes) to work around this, though I hear rumours that dry hair shampoo works too that I will test because scanning spray is 35 quid a can for the aesub blue.
So Its not metrology grade despite their claims and the software is a bit buggy and it needs a really fast high end pc, but within 0.2mm is enough for the vast majority of work, building fixtures to hold things, car trim, stuff you will 3d print anyway because well, its plastic squirted out of a computer controlled plastic glue gun. Maybe as I get more seat time on it, I can get nearer that 0.1mm as it definitely is a skill you have to aquire.
Your not going to use it to align something to micron accuracy, but it is what it is.
There are some programs that can mesh the point cloud and extract geometry for a traditional vector drawing of a part, but still investigating and learning that as a vector cad program is more use than a point cloud in a lot of circumstances.
Posted elsewhere here already, biggest scan Ive done so far, a rover v8 4.6L block. The orange areas represent insufficient data, but that was because the scan spray subliminated off the liners/inner crank web area before I got to them, and if you re-apply it blinds all the marker dots making tracking & scanning impossible. Its perfectly adequate to make some fixtures and for designing engine mounts, exhausts etc in future from.
I've scanned and 3dp some smaller things direct from the point cloud model exporting to stl then slicing it already, a replacement hydromassage jet for the wife's spa and the funny little twist to lock fixtures that hold the end fuse panel cover on the slk dash as some were missing.
You can scan in just point data (the rover block), or add a scan taken from the visual image (like in the slk console scan which shows the marker dots up visually), the rover I didnt need visual clues and it cuts down on processing size, its already at nearly a million points in the cloud with just raw data without image mapping.
Post processing I spend some time manually removing stray dots/noise which all photogommetry suffers from to some degree (including the 100k industrial solution my son uses in his job as a industrial metrology technician...), then rubbish removed, I proces it into a point cloud and mesh it, then if there's some areas not well defined or to scan from underneath etc (the rover v8 is a true 360 model, you can rotate it round in 3d in cad and see into it from the sump etc) take more scans and add them the project. Then I pick markers on the two scans that are in two or three planes & show up in both scans and use these to align the scans. So I put markers & I have some tall marker targets on magnets to "walk" around corners to get a extra plane of data for this. Feature alignment or non marker mode is a joke, it always messes up, but manually guided by a mk1 eyeball its really accurate and works well. Once you have them aligned the software asks if its ok, and you merge it into one point cloud model. Then you can take more scans, and repeat adding them the model each time, or in one giant mergathon at the end if you prefer.
I think the rover block in the end was 10-12 separate scans merged into one model but that was a complex awkward thing to scan, normally I take 2-3 scans of something to scan it in 3d. If I needed the liner detail I could always go back and mask off the tracking dots and spray it with scan spray and repeat to eliminate the orange areas, but its not needed.
The slk plastics were a nightmare to scan at first, but once the "soft touch" plastic spray coating put on at the factory had been removed (which was needed anyway, it had turned to sticky snot with age like a lot of merc/bmw interiors of the era), they didnt need spray to work apart from the odd bit I missed a bit of soft touch coating on somewhere hidden where a quick sniff made life easier.
I spent about 3 weeks just 3dprinting marker towers with my resin printer and sticking target dots on them and gluing neo magnets into them, going cross eyed and used 2500 marker dots but a worthwhile investment in time I think, and they are suprisingly expensive to buy. You can see some of these stuck the shelving in the background in the actual photo of the rover block on the bench.
I havent used it in IR mode, that seems a bit playtown toy for scanning faces with no reference geometry to me.
@doubleboostMmm, that might be an option.
However, if I am going through the trouble of casting, is it not better to then just cast them in metal?
Who are the casting boffins on this site?
sako243
Member
- Messages
- 4,071
- Location
- My mansion in Wales
Surely @pressbrake1 is the casting master. Well at least makes the biggest.Mmm, that might be an option.
However, if I am going through the trouble of casting, is it not better to then just cast them in metal?
Who are the casting boffins on this site?
