Video Probe Technical Discussion - was WTB small xyz motorized stage
- Hello Henry,
Some comments about my 90 degree preference on the forum.
As far as building a fixture...this is a bunch of machinists. Surely not an overwhelming task if drawings are provided. I will eventually get around to posting my drawings too.
BTW, can you be more specific about your laser and camera setup. Can you post some point clouds or other results or photos on the forum?
More later. Gotta run right now.
>From: Henry Liu <henryjliu@...>
>Sent: Jun 1, 2009 4:41 AM
>Subject: Re: [mach1mach2cnc] WTB small xyz motorized stage
>I read through some of your posts and it looks interesting. From your
>Correct me if I am wrong but it seems you require a very carefully
>machined/aligned created setup to create the 90degree angle?
>If so, I do not think this would be the best solution for a
>In my setup, I use a 1/4"-20 threaded nut and superglued it to the
>back of my Quickcam 9000 camera.
>I then bought a RAM mount with 1/4"-20 bolt and attached the whole
>thing to my gantry using bolted down 80/20. I then took my laser line
>and mounted it at an arbitrary angle with a bolted hose clamp.
>There is almost no information assumed on the camera and laser
>geometry. Totally assembly time is 10mins with off the shelf
>components. I did not machine a single part.
>Instead, I wrote an advanced calibration algorithm where I image 10-20
>different views of a chessboard. I use corner detection to extra the
>points to 1/10 pixel accuracy. I calculate the rotation + translation
>matrix to translate the chessboard reference frame to the camera
>centered reference frame. I also take into account lens distortion
>and other factors.
>I then shine the laser on the chessboards to collect a bunch of points
>on the planes. Since the planes are not coplanar, the intersection of
>the laser plane and the chessboards gives a bunch of points that lie
>on the image plane.
>I then least square fit the laser plane.
>Now, whenever I see a bright spot pixel (h,v) in the camera view, I
>have a lookup table that tells me the Z depth.
>Due to the calibration technique, the geometry between the laser and
>the camera is unknown at construction but calibrated individually.
>This minimizes the cost in my opinion by using more complex software
>to make up for poor construction tolerances. If you bump the
>laser/camera out of alignment of course you need to recalibrate
>everything so use lots of superglue :)
>On Sun, May 31, 2009 at 8:16 PM, Tom Hubin <thubin@...> wrote:
>> Hello Henry,
>> I seem to recall your mentioning least squares fitting for calibration. I
>> cannot find that email right now but here is a link to my method using a
>> single point target viewed from multiple positions. Least squares fitting is
>> used and I am pleased with how well it works in the Scan3d plugin for Mach3.
>>>For lightweight laserscanning, why not just mount a mirror on a stepper?
>> Thanx for the suggestion but scanning is not my goal. I am doing this to
>> keep my brain from turning into mush while I am pretty much homebound. I
>> designed and built non-contact measurement equipment in the 1980s. I want to
>> use my expertise to give something back to the CNC community that has
>> steered me into fun and education for the last 9 years.
>>>People have already pushed the limits of laser scanning in a field
>>>called confocal microscopy:
>> I am familiar with confocal microscopy. Designed some experimental equipmnet
>> for NIH in the late 1980s. Used an acousto-optic scanner for the laser and a
>> scanning pinhole IDT (Image Dissector Tube) for the image. Produced
>> interlaced 60 frame/second video rate confocal microscope. Viewed on a
>> standard monitor. Published papers are out there somewhere.
>> But this is way to pricey and complicated for the DIY CNC market. I think
>> simple video probing with a webcam and a laser line generator module is much
>> more viable for DIY. Some education is needed so that folks can setup the
>> hardware properly and get reliable and accurate data. Then, IMHO, 3d video
>> scanning will become popular.
>>>Hit me up if you have any questions. I've had a lot of experience
>>>with it and read many papers on the subject.
>> I look forward to some technical discussion but we may have to move it off
>> list if we get too far afield. However, as long as we are discussing methods
>> that use Mach3, or can use Mach3, there will be some benefit to the list.
>> Please feel free to comment on my Open Source code or any of my comments
>> here or in the Mach3 forums.