Improving my lab performance by digital transformation I want to develop a software which perform a calibration of an analog instrument. Advances in computer vision with Pyton addresses the stuf but fot me it has the error that the reading of the gauge depends on angle of the needle, not properly over the scale. So if the scale is not correctly printed that bias will not be detected by the vision. ¿Do you know any alternative?
you could train a supervised model to learn that bias, if you show it enough images with the needle off, and what the ground truth value should be
Dear Pablo Constantino ,
Look the link, maybe useful.
https://medium.com/@nayak.abhijeet1/analogue-gauge-reader-using-computer-vision-62fbd6ec84cc
Regards,
Shafagat
Pablo Constantino ,
Is there a reason it has to be by computer vision? Why not couple the gauge needle to a potentiometer on a small circuit that is connected to a Raspi?
Just out of curiosity
Alternatively, if you need to use computer vision , then I would suggest a strong light source on top of the gauge so that you can properly calibrate the angle of the camera against the shadow of the needle. I would then take careful directed measurements to calibrate the angle of the needle from its start position to its current position and translate those into pixel offsets that can give you the angle based on pixel differences.
Hope this helps
Update:
If the gauge is too small you can still use the shadow of the objects (which you can mark on the background) to calibrate the angle and position of the camera with respect to the gauge.