Which programming language should I try and learn to translate a lines visualization in a screen dump to an expression of “variance” compared to a straight line - I hope R or Python but am open to other suggestions - but if you think the below challenge can be solved using R or Python - I would love to jump into the learning!!
My task is to be identifying a colored line that I know is within a picture/screendump - and translate the plotted time-serie line of the depicted line into an expression of variance? (how many and how big variations in the line progression from the green line appears until it stops.)
I would like to understand which library/package/accellerator) you think I should pursue to understand to solve this…
I have an output screen in an old system where I cannot read the source values - I can only do a screendump of the status screen manually in. I would like to get to the actual values. The output Log screen shows a diagram area where a green line is illustrating machines recorded output values over time since the batch was initiated it is an unbroken “non-linear” line where the value slowly decreases or increases over time alternating from e.g. 30 minutes of decrease to 5 minutes of increase to 10 minutes of decrease to 20 minutes of increase etc . Currently we just look at it manually to evaluate if we need to clean the valves. I would like to be more facts based.
If the green line is swinging a lot, the system needs service. In other words the deviation of the green vertical position of the center of the green line compared to a simple Y=aX+b So, which language would you choose to process the image. I imagine I will do the following:
- Crop the image (I know which area is the plotter-area of the image in which the line is identifiable
- Identify where the green line begins (e.g. from searching from the left of the cropped picture until I find one of the RGB values of the lines composition (it is “static imitating a glowing line” so the RGB values are 12 different nuances of green)
- somehow analyze how the line is fluctuating - my idea is to divide the 2000 pixels to steps of e.g. 100 pixels - identify the green lines vertical position at the first and last pixel of each segment. Calculate the formula (Y=aX+b) matching start and end value of the segment, and the traverse from left to right inside the 100 pixels segment to find the green line and see if it deviates from the straight line of the 100 pixels and sum the deviations.
If someone can help with a smarter way to calculate the area between a straight line and the flow of this green line, I am all ears.
Looking forward to hearing you geniuses help me out /Søren - a.k.a. Handstand