IDEA #8R0R9R Deep Learning Approach for Data and Computing Efficient Situational Assessment and Awareness in Human Assistance and Disaster Response and Damage Assessment Applications. 23A0022

The state-of-the-art method to process and analyze the matching pre- and post-disaster/action images are contrastive learning (not an invention per se, but a current popularly used deep learning method in machine learning), where the two matching images are learned via the same Siamese network to extract representations. The new representation developed in this work is an important breakthrough since we encode the difference and importance of the target of interest region from the 2 matching images into a single image, in consequence there is no need to limit ourselves to contrastive learning to deal with the 2 images, but using any deep learning methods for single image segmentation and classification, this way the learning performances can be improved significantly. The work will be of practical use to Air Force mission such as battlefield assessment, i.e., checking the situation of a regions of interest for action evaluation. Furthermore, this work can find ready use for government agency or safety business to better evaluate situations for action planning.