Disaster relief using satellite imagery

We have satellite images of disaster stricken areas such as refugee camps taken at different times. We want to analyse or compare these images to see the progress of disaster relief operations or to measure level of severity in that particular area over a period of time. For example, if we have satellite images of refugee camps, we want to count shelters to see if the number is increasing or decreasing. In these days, UNOSAT has staff dedicated to analyse and then classify objects in these images manually. This is an arduous task, takes a lot of time, inhibiting quick reaction from relief organizations. Conventionally, objects in images are identified by using color or shape features. But in our case, all objects such as shelters are arbitrary in shapes and color. Sometimes the background surface has the same color as the shelter. So, it is not possible to employ any conventional color or shape based image segmentation method. The solution we are proposing is "Automated Feature Classification using Machine Learning". My team at CERN openlab with UNOSAT are evaluating different machine learning based feature extraction algorithms. Currently, we have evaluated Facebook's DeepMask and SharpMask object proposal techniques