I don't know what kind of objects do you want to detect from this images (if the
"Automatic Extraction of Exposure data" is a known term), but in the kaggle website I found a competition which is titled "Dstl Satellite Imagery Feature Detection" and there are some kernels which are free (https://www.kaggle.com/c/dstl-satellite-imagery-feature-detection) a good approach could be found here (https://github.com/toshi-k/kaggle-satellite-imagery-feature-detection).
Depending on which kind of objects are going to detects the developed method would be different, this is another example of object "Object detection on satellite images" here (https://github.com/marcbelmont/satellite-image-object-detection).
I think it it too hard to develop an object detection in general without knowing about object's type or even database.