search

TinWhiskers / fpd-design

Design document for FirePick Delta
1 stars 0 forks source link

Determine camera calibration transformation matrix values for whichever cameras+lenses go on the final DFM FPD machine #44

Open FlyingLotus1983 opened 9 years ago

FlyingLotus1983 commented 9 years ago

This may require making a fixture to mount the cameras into. This fixture may be used at a later date to measure all cameras at the factory, either as a go/no-go type, or, possibly as calibration parameters (i.e. each machine may have slightly different params according to the individual camera variations).

This data will be used by OpenPnP to un-warp the image.

FlyingLotus1983 commented 9 years ago

Assigning @vonnieda to this issue. We will need matrixes for the 5 megapixel camera (fixed focus for now, possibly auto-focus?), and the ELP 100USB. We might change the ELP 100USB over to a 5 megapixel for the up-looking camera, but that's an open issue for now ( https://github.com/TinWhiskers/fpd-design/issues/66 ), so the highest and first priority is to get the 5MP OV5640 module on the end effector mapped out.

vonnieda commented 9 years ago

Just a note - already did this for the setup I am currently using (5MP OV5640) but it will have to be re-done if we change the focal distance.

We're also going to need to come up with a process for either calibrating every camera/machine or setting the focal distance to something specific before shipping a new machine. The camera calibration matrix is specific and sensitive to focal distance.

On Mon, Oct 12, 2015 at 11:05 PM, Neil Jansen notifications@github.com wrote:

Assigned #44 https://github.com/TinWhiskers/fpd-design/issues/44 to @vonnieda https://github.com/vonnieda.

— Reply to this email directly or view it on GitHub https://github.com/TinWhiskers/fpd-design/issues/44#event-433040092.

FlyingLotus1983 commented 9 years ago

OK. For the production machines, it should be doable to make a jig to hold the end-effector and camera assembly to do a calibration of each and every camera that goes out the door, assuming the software app that performs the calibration is reasonably stable and repeatable.

vonnieda commented 9 years ago

Yep, that's doable. I also think it would be easy to have users do it or maybe even do it automatically.

And finally, if we can find a way to set the focal length very accurately we should just be able to reuse the camera parameters.

On Tue, Oct 13, 2015 at 11:04 AM, Neil Jansen notifications@github.com wrote:

OK. For the production machines, it should be doable to make a jig to hold the end-effector and camera assembly to do a calibration of each and every camera that goes out the door, assuming the software app that performs the calibration is reasonably stable and repeatable.

— Reply to this email directly or view it on GitHub https://github.com/TinWhiskers/fpd-design/issues/44#issuecomment-147582263 .