Multimodal and Multi-Temporal in Satellite Images Based on Flood Detection: A Comparative Study

Floods can cause significant damage to human lives and property, necessitating precise post-flood evaluations for rescue efforts. Methods: It explores flood location AI techniques using a multi-modular and multi-fleeting picture dataset, ranging from traditional methods like multi-facet perceptron (MLP), support vector machine (SVM) and Deep convolution neural network (DCNN) to space transformation-based strategies. We utilized SPOT-5 and RADAR pictures from the November 2000 flood in Gloucester, UK. Results: The F1 score of 0.8846 and the region under the accuracy (AUC) bend (PR) of 0.9173 in the examinations were marginally higher than those that used machine strategies and investment of human resources. Conclusion: Utilizing a multi-modular and multi-temporal image dataset looked at a few customary flood location AI techniques as MLP, SVM and DCNN, to later space transformation-based strategies. This enhancement is due to the 20 labelled data samples used in the domain adaptation- based on semi-supervised domain adaptation (SSDA).