I have a number of images, (these could be vectorised) that are label classifications. There are only 2 classes, 1 or 255. Each image is from a different date. For each date I also have data that can be in csv or point form.
I wanted to try and use logistic regression to get a future, binary prediction using the images, as the dependent variable and the point data associated with each pixel as the independent variables.
I've done this with one pixel, I need to try and extend the model to a wider area and I'm having trouble visualising how to do this. I had thought of having it all as numerical data and looping through the process for each pixel but this would take a very long time.
I am not too familiar with GIS so I could be missing what's possible, but I had also thought of converting the raster to points and populating the attribute tables for each point by joining tables by geolocation. I could convert this values in to rasters as well using spatial interpolation.
I believe there's also some methods in R or python. Or maybe there's some data fusion method I'm missing. I'm a bit stumped on the matter so I'm looking for ideas on how this might be done.
Maybe someone has experience with a similar problem?
Basically I want to input images and text data/point data in to the model and have my output as a predicted image. Maybe this is impossible in GIS.