Week of 11/26

This week was spent on design and preparation for the document.

Tyler and I met last Wednesday and we made a preliminary class diagram.

This Tuesday, we continued working on more design elements.  We have revised our class diagram, especially after meeting with Dr. Wolff and it is still a work in progress.  I have also done state diagrams for intrinsic and extrinsic calibration as well as scanning.  The calibration ones have changed based on MVC design and I will edit the Scanning one as well.  I feel better about our MVC design after Thursday's work and discussion with Dr. Wolff.  All in all, design is moving along fairly well.

Grady was able to do some of the hardware and hardware-related design charts as well.  He also got a book from Tosh about circuits that has a chapter about interrupts that should prove useful.

We present on Tuesday; so Tyler and I are going to meet up on Saturday (Grady will be out of town) to prepare the slides.  We will correspond with Grady to get some of the hardware information and will send the slides out so Grady can also review.

Other terms of business:

• The Vec3F structure of OpenCV does have the cross product function so we should be able to use it for our vector functionality.
• We found the fitLine function in OpenCV but were unable to find an approximate intersection function for 3D lines.  We may have to write the algorithm for this.
• Tyler looked into possible ways of integrating GUI design with OpenCV; there is a C# wrapper available but most people suggested Qt, which is likely the route we will pursue.

We should be on a good pace to supply a quality design document due in a little under 2 weeks.  We have a good design foundation and will continue making progress.

Once design is done, we will be on break and be able to come back ready to start implementation in J-Term and the spring semester.

Week of 11/19

Short week this week with Thanksgiving.  But that didn't stop us from working :)

Tuesday we met to focus mainly on design.  Tyler and I were able to work out state diagrams for intrinsic and extrinsic calibration as well as scanning.  Tyler also continued work on the document itself in other areas, such as research review and testing plan.

Grady had a major breakthrough: he got the motor spinning.  This was a great step to achieve, especially right before the holiday weekend.

We met with Dr. Wolff and discussed some things, including reviewing our requirements document briefly and discussing design, namely the classes needed.  Tyler and I came in on Wednesday and worked on figuring out the classes we will likely need as well as some methods/pseudocode for what the classes will do.  Tyler worked on the UML diagram and has a preliminary.

Overall, it was a pretty productive short week.  We will come in next week and continue working on the design document and prepare for the presentation (maybe next Saturday??) which we will give the following week.

Happy Thanksgiving to all.

Note: Here is what the plot of the red component, not intensity, looks like for the laser line:

Week of 11/12

Another week gone by...

This week was rather so-so as far as progress.  Last Thursday, Grady and I started configuring our machines.  This took a significant amount of time, but we now have 3 machines (2 Windows, 1 Ubuntu) to work with.

Tuesday, I also researched the undistort/z-component and determined we should undistort (since it uses 2D) and then create a homogeneous coordinate from the idealized coordinate.  Speaking with Dr. Wolff, he still wondered about what the undistort function did with the z-component.  So Thursday, I took the example calibration code, used the pictures provided, got the intrinsic and distortion parameters, and then plotted out the distorted and undistorted points.  It took quite some time, especially with the example code in C.  I tried to convert things to the more recent C++ openCV elements, but I ended up sticking with C (the new version of the OpenCV book comes out in December; hopefully it includes updated example code we can use).

Below is the comparison of the distorted points with the undistorted (in ideal camera coordinates).  They seem to look pretty good and I think we can go ahead and proceed using the procedure described above. (thus the z-component must be assumed to be 1).
Note: the image used is 1600x1200 and the scaling between the two is off.

On Tuesday, I also was able to use just the red component of the laser (rather than the intensity level) and the data did show a spike.  I tried in a light room as well and didn't get very good results.  More testing in a light room with better camera settings will likely occur later.  But the dark room will produce the most stark results.

Grady is still chuggin' away with the circuits, communicating with the company.  He seems to have gotten a little more of a grasp and hopefully is close to a breakthrough.

Tyler researched some on the opencv video settings.  He hasn't been able to meet recently on our normal days, so hopefully he is available more as we approach the end of the semester.

We have a design presentation coming too quickly and will likely need to turn more of our focus towards that.

UPDATE:

Here are the mappings of the corners to the distorted and undistorted images of a chessboard:

Distorted image

Undistorted

Week of 11/5

I'd call it a successful week.

Monday, I read through the entire math section in the Siggraph document. Parts of it took a while but I feel I was really able to understand almost all of it, including partially understanding the partial derivatives.  The partial derivatives are just used for minimizing to find the optimal approximate intersection between two lines.  This likely won't really need to be a focus point in our document/presentation.  Anyways, back to the other stuff; I am glad I re-read it.  In fact I wasn't even sure I ever read the coordinate system part.  It helped to show a piece we had missed as far as the inverse intrinsic matrix goes.

We met on Tuesday primarily to finish up the mathematics pipeline.  It started out fairly rough once we hit solving for lambda.  It took some time and was frustrating.  But then I looked back at the document and found a way to do it, clearing a main hitch.  From there it was fairly straightforward.  One main question remaining is the z-component of the real image coordinates.

We met with Dr. Wolff after the tech-talk and I walked through the math.  He seemed quite pleased and offered an additional piece we had missed as far as the distortion parameters went.  I feel from last Tuesday to the end of this Tuesday, at least for me, I now understand the main processes much better.  Obviously it is still in mathematical-type pseudo-code, but it is a critical foundation we can build on.

For our next meeting, we are working on a few different things, namely some of the following:

• More accuracy for red plot: taking the R value from the BGR rather than the V value from HSV
• Red plot location accuracy (coordinate rather than a pixel value); how to better determine point (left side of laser?)
• Motor/circuitry (GRADY)
• LaTeX for the math pieces (JEFF--Done Thursday 11/8)
• Settings for camera from OpenCV
• OpenCV undistortPoints method
• Using K^-1 and Distort parameters
• What about the z-component of our image?

Grady has hit a bit of a hitch in the circuit code compilation from within Visual Studio.  But hopefully he can figure it out; we have Windows installed on two machines now and will work on getting them configured more fully.

Again, Tyler is at a disadvantage at this point as he hasn't taken linear algebra or graphics, but he did read (and take notes) on at least a good chunk of the Siggraph math portion and worked on learning some basic matrix calculation procedures.

All in all, a successful week.  Things are lining up nicely.  Design document looms in the future.  And we are on a good track...

Week of 10/29

Well the requirements document is all done :)  This week we focused more on the math and the intensity plot.

Tuesday, we spent some time on the intensity plot.  I had been overcomplicating it and had been looking at the opencv histogram function.  However, we determined to pursue by looping through a row of pixels (the row is just halfway down the image) and getting the intensity.  We tried using the red frame piece we had previously generated, but the channel we were testing had only 0's.  I think the only channel is the 1st (s[0].value) but it would only contain a 0 or 1 depending on if the hsv is in the range we specified.  We then tried with hsv and got some data.  Tyler furthered it to an excel plot and Thursday we plotted the laser line in a dark room.

Tuesday was also spent in a lengthy meeting with Dr. Wolff discussing the mathematics and starting to work out an order of the operations we will be performing.  I wrote out some things Wednesday morning on our board from what we had discussed and Thursday we worked through some of it.  We were able to Skype Grady as he was working at the time.  This was a way to communicate without him actually in the room.  We struggled some (especially on how to determine the coordinates of the laser line when determining the laser plane, until finally figuring out that the line lies on a plane and thus can determine the world coordinates in that plane's coordinate system).  Tyler is at a disadvantage not having taken linear algebra, but I am trying to explain some things as we go so he isn't lost as much and can feel more included.  We still have some math to figure out, particulary the ray to a point and the ray-plane intersection to determine the coordinate in one of the coordinate systems (both the ray and the plane must be in the same coordinate system when doing the intersection).

All three of us are going to review the math portion of the Siggraph document.  I read through this already, but it has been some time ago and it will be helpful to refresh.

I think we are making headway and I hope we can keep plowing through.  It sounds like Grady has a way to move forward on the hardware side of things as well.