Dimension measurement

[muks]

can you guys share your application result on the minimum dimension measurement acheived in labVIEW. I understand that it depends on lot of parameters like FOV,Working distance.resolution,lightings etc.But still i have acheived measuring +\- 70 microns for measuring a contamination consistently.

[Eric_R]

Hi Muks,

I have heard of readings less than +/-20 microns read. You can potentially go higher and higher if you used a camera based off a microscope. As cameras increase in resolution we will be able to see more and more accurate images. Given proper lighting the Equation for the size of the smallest feature is "Size of smallest Feature = (FOV/Sensor Resolution) x 2". Let me know if you have any questions.


Thanks,
Eric Reid

[muks]

Hi eric,
           You are right.But neglecting all this and if i go for sub pixel sampling which i am working on i should prove the formula wrong right?Like if  sample each pixel to 10 sub pixels.I know it is easier said than done.Have tried something on this line?

[BruceAmmons]

Using a microscope, I was able to measure at the 1 micron level.  I think we got pretty close to 0.5 micron pixels.  However, my field of view was only 500 microns or so.

You can use sub-sampling to measure smaller increments than the pixel size, given that your image has high contrast at the edges.  However, the object still needs to be about 2 pixels wide.  If the rising and falling edge occur in the same pixel, there is no way to determine where the edge would be using sub-sampling.

Bruce

[muks]

---

I think we got pretty close to 0.5 micron pixels

---

Were you able to acheive the repeatability?

[Eric_R]

Hi Muks,

You are correct the formula I posted does only represent pixel accuracy and not sub pixel accuracy. For sub pixel accuracy basically you are performing a curve fit to estimate where the actual edge is. With the imaging setup and components currently on the market you can reliably estimate up to 1/25 subpixel accuracy. The results you receive from subpixel calculations are very dependent on your image setup. For further questions as to the equations used in the edge fitting process please reference section 11-9 on the NI Vision Concepts Manual. Keep this thread going it’s great to see the full power of NI Vision put to the test!


Thank You
Eric Reid

[muks]

Hi Eric,
           But still you need atleast 3 pixel to resolve an edge right?So does it mean even when we go for subpixel accuracy we could atmost analyize the exact edge for one image and cannot acheive repeatability for real time images?

[Eric_R]

Muks,

It is my understanding that the algorithms used in sub pixel accuracy interpolate between two pixels on both sides of the pixel that the interpolation will be performed over. The peak calculated position is determined to be position to the nearest whole pixel.  As far as reproducibility goes, if the real time image is the same it should give the same results from the sub pixel math, however any minute changes could affect this.


Thanks and have a great 4th of July Weekend
Eric Reid

[muks]

---

however any minute changes could affect this.

---

That is my greatest fear.I think i will end  up  with analysing something rather devaloping something for building a machine.

---

Thanks and have a great 4th of July Weekend

---

What is special abt july 4.We are working on that day.

[MikeS81]

---

muks wrote:

...
What is special abt july 4.We are working on that day.

---

Hi muks,

july 4. is independence day.

Mike