Advice needed on substrating a reference frame from 15000 jpgs

[Bruno Schroder]

I'm still busy with the processing of the data and images from the Microclimate n Bat Roosts project. Massimo's answer to my previous post is now in production to build presence maps, like the one below.

My next problem is to plot the number of bats through time. For that to happen, the first step is to substract from each picture a frame with no bat in it. As the camera is fixed and IR illumination is constant, I'm then left with blobs, representing the bats, which give an approximate count of individuals when  the colony is spread accross the roost. Not perfect but much better than counting them manually on 15.000 pics per roost, knowing next year's campaign will deliver over 100.000 images

Would you know a software that would do this in batch? Like substract a specific frame from an entire directory of pictures (JPG)?

Project background:
I'm involved in a multiyear bat conservation project trying to measure the influence of microclimate conditions in occupation of bats roosts. I'm in charge of the data acquisition, meaning temperature and humidity sensors and IR recording of the presence of bats.
Roughly speaking, I'm getting 600K datapoints and 100K images a year. Images are reviewed by volunteers and when bats are present, images are tagged and a pointer is added to the database of the sensor data at the right time and date. In a second stage, data scientists will look for correlations between whatever interior or exterior climate data and the presence of the bats. We are trying now to extend the amount of information by counting bats over time.

[Alaun]

Bruno, your problem sounds similar to what is done in metallography, when particles or phases have to be count in an alloy. That is done -when not manually- with Image analysis software. I only know some professional software, which is very expensive (IMAGIC). But I have googled and it seems, there is also free software available to do this. E.g. I found this: https://imagej.net/  Perhaps that is helpful to find something for your purpose.

Here is a further link right into the problem of counting: https://imagej.net/imaging/particle-analysis


And some other software, called ORBIT:  https://www.orbit.bio/tissue-quantification/
You must replace "Tissue" with "bats" ;-)  This software does not like jpg-pictures, they have to be transformed into some special format.   

[Bruno Schroder]

Werner, I hadn't thought about metallurgy. Indeed, it is similar., with the additional burden of the large number of pictures to handle.
There's lot of good stuff to investigate in the two software you reference. That's very usefull.

Thanks a lot, Werner.

[Jack Dahlgren]

I work for a company which does a lot of video analytics and also AI development. If you could share some samples and a statement about the value of the project I’ll see if I can find someone who would be interested in doing a little side project.

[Bruno Schroder]

That would be great! I'll send you a separate document with samples. DM sent.

Thanks a lot, Jack

[JCDowdy]

You might want to take a look at ImageJ https://imagej.net/downloads or Fiji https://fiji.sc/, both free and open source.
If you are lucky you might find a macro/script that you could adapt to your project.

[Bruno Schroder]

Woaw, the number of plugins on Fiji is overwhelming ... Thanks for the link, John

[Alaun]

A kind of a tutorial/intro:

https://www.southampton.ac.uk/~assets/doc/Image%20Analysis.pdf