Hi, I am trying to develop an application to detect frost/ice on commercial air conditioner’s evaporator coil. We can install cameras on site to take pictures of the coil periodically. My goal is to have a rough estimate of how much frost/ice is formed (say 30% of the surface is covered with ice).
I am a programmer with years of java/C++/python experience, but I have never used OpenCV.
Is OpenCV the right tool for the problem I am trying to solve? Can someone point me to a book/video tutorial to get started?
I don’t think looking is the right solution. This kinda stuff ought to be solved with simple sensors. the ice can conduct electricity, right? then you just need a bunch of exposed non-corroding wires, and when the ice closes the loop, you’ll know.
If they do it now by looking, it is the right, proven solution. They just need to automate the tested procedure and not do any sensor system development and tuning.
you’re neglecting the step from humans looking at the ice to a machine looking at the ice.
I think the development cost and installation cost for computer vision – and this will likely require AI – will at least match something based on simple electronics.
a quick google search throws application notes from Texas Instruments at me. five-year-old application notes. and a reference design that you can just take and copy and do with it what you want.
I think a talk with the manufacturer of those HVAC units is warranted. they better have a solution for this.
this isn’t even a “solution in search of a problem” because the solution, making a computer recognize ice/frost in a picture, isn’t even done yet. electronic sensors clearly are.
wow, thanks for the replies. Really appreciate it.
Actually I am working for a software company that provide remote management solutions for HVAC manufactures, and yes their units already have frost sensors. However quite a few of them, after deploying our software, ask if we could integrate our software with a camera, for visual frost detection. And hence the problem.
So in this particular case, we already have a remote management software system, and we are already working with manufactures to collect remote data. The only missing link is auto detection. (We are not trying to propose an AI solution from scratch).
So perhaps I should rephrase my question like this, if we were to use software to automate frost/ice detection on coils, is OpenCV the right tool? If so, which algorithms in the OpenCV library? (I assume this is a very simple problem for computer vision. After all, if cars can auto drive, how hard can it be to detect ice from static images).
a customer requirement… should have guessed. they’re always good at coming up with entertaining things to do.
OpenCV is probably a good choice if you’re doing image processing, computer vision, or AI inference.
to finetune an AI model or train your own AI model, you’d use pytorch or tensorflow or maybe matlab or whatever else is out there.
the problem can be very hard, depending on how hard you make it for yourself. you should go with AI and let the AI learn instead of coming up with detection methods on your own.
you will need data, plenty of data.
and you’ll need a decent background in all of this to have the judgment to know how to prepare the environment and place the cameras.
if the environment is locked down, lighting-wise, stuff could be easier. it’s probably not though. go with AI.
