Improving Screen Caliper Accuracy
Synopsis: In the prior article, Real-world Measurements in Virtual Space, I introduced the concepts of using virtual calipers to measure real-world objects on your computer screen. In this article I show how to improve the accuracy of the measurements you take on the screen by carrying forward the same example of measuring various dimensions of a small printed circuit board. It's possible to greatly improve the accuracy of Screen Calipers by wisely choosing your calibration points.
In the first video of the previous article, I showed the calibration process by calibrating using the known spacing of the breakout board pinouts. Because this board has standard 0.1" pin spacing, I calibrated based on the outside edge-to-outside-edge of a pin pair that represented what I knew to be 0.1". This is a good start but can be improved. Here is a demonstration of the percentage errors and improvements in accuracy based on better calibration.
First, here's the actual caliper measurement of the physical board in order to show what the goal is. Obviously, if you have the physical board and calipers, you wouldn't necessarily need Screen Calipers, but I'm using real calipers to compare to Screen Calipers in order to illustrate how accurate Screen Calipers can be. The only way to do that is to compare actual measurements or an engineering drawing.
These measurements are inches because I calibrated to 0.1" and wanted to keep the units the same throughout the illustration. So, the actual width is 1.35 inches according to my real-world calipers. The height is measured:
and is 1.105 inches.
Compare these actual values to the initial Screen Caliper measurements when the calibration was done on a single 0.1" hole spacing:
This shows 1.424 vs the real 1.35 inches - off by 7/100ths of an inch. Doesn't sound like much, but it's quite a bit and we'd like to do better. The height was measured by Screen Calipers to be:
1.156 inches versus the actual 1.105 - off by a little over 5/100ths of an inch.
Instead of calibrating to a single 0.1" hole spacing, I calibrated to 10 holes or exactly 1" on the board. After setting the calibration to a larger length of known dimension, it's possible to get very accurate readings. After the new calibration, here are the results:
Screen Calipers shows 1.352 vs the actual 1.350 - we reduced the error from 7/100ths to 2/1000ths of an inch just by picking a better calibration point pair. The height compares like this:
Screen Calipers reads 1.098 vs the actual of 1.105. This reduced the error from 5/100ths to 7/1000ths...again a great improvement.
The reason calibrating on a smaller scale, even if the distance in the photo is well known, is that there are both accuracy and precision errors in setting the length and that error is a larger percentage of a small distance than it is a large distance. So, by calibrating to a small distance and then measuring a larger distance, you multiply that error over and over across that larger distance. By calibrating to a larger known distance, that error is a smaller portion of the distance and therefore any error is much less when measuring smaller or similar distances.
Conclusion
The moral of the story is that you can get highly accurate Screen Caliper measurements by calibrating to the largest known distance within the subject photograph.
In the first video of the previous article, I showed the calibration process by calibrating using the known spacing of the breakout board pinouts. Because this board has standard 0.1" pin spacing, I calibrated based on the outside edge-to-outside-edge of a pin pair that represented what I knew to be 0.1". This is a good start but can be improved. Here is a demonstration of the percentage errors and improvements in accuracy based on better calibration.
First, here's the actual caliper measurement of the physical board in order to show what the goal is. Obviously, if you have the physical board and calipers, you wouldn't necessarily need Screen Calipers, but I'm using real calipers to compare to Screen Calipers in order to illustrate how accurate Screen Calipers can be. The only way to do that is to compare actual measurements or an engineering drawing.
These measurements are inches because I calibrated to 0.1" and wanted to keep the units the same throughout the illustration. So, the actual width is 1.35 inches according to my real-world calipers. The height is measured:
and is 1.105 inches.
Compare these actual values to the initial Screen Caliper measurements when the calibration was done on a single 0.1" hole spacing:
This shows 1.424 vs the real 1.35 inches - off by 7/100ths of an inch. Doesn't sound like much, but it's quite a bit and we'd like to do better. The height was measured by Screen Calipers to be:
1.156 inches versus the actual 1.105 - off by a little over 5/100ths of an inch.
Instead of calibrating to a single 0.1" hole spacing, I calibrated to 10 holes or exactly 1" on the board. After setting the calibration to a larger length of known dimension, it's possible to get very accurate readings. After the new calibration, here are the results:
Screen Calipers shows 1.352 vs the actual 1.350 - we reduced the error from 7/100ths to 2/1000ths of an inch just by picking a better calibration point pair. The height compares like this:
Screen Calipers reads 1.098 vs the actual of 1.105. This reduced the error from 5/100ths to 7/1000ths...again a great improvement.
The reason calibrating on a smaller scale, even if the distance in the photo is well known, is that there are both accuracy and precision errors in setting the length and that error is a larger percentage of a small distance than it is a large distance. So, by calibrating to a small distance and then measuring a larger distance, you multiply that error over and over across that larger distance. By calibrating to a larger known distance, that error is a smaller portion of the distance and therefore any error is much less when measuring smaller or similar distances.
Conclusion
The moral of the story is that you can get highly accurate Screen Caliper measurements by calibrating to the largest known distance within the subject photograph.
asdfasdf
