Deep Disaster-Net: A Multi-Objective Gradient-Hopping Optimized Framework with Adaptive Classifier Fusion for Post-Disaster Image Segmentation

Abstract

Analysis of satellite images is an essential role in the response and recovery of disaster situations especially after a natural disaster like the hurricane. The conventional approaches of assessing damage heavily depend on manual examination and the antique paradigm of classification having the lexicon of limitations of being scalable, versatile and expedient. This paper presents the hybrid machine learning framework to focus on the problems of post-hurricane satellite images segmentation and classification. The solution being proposed is one which interpolates a Gradient-Hopping Hybrid Optimizer (GHHO) and a Pattern-Adaptive Classifier Fusion (PACF) mechanism. GHHO amalgamates, global exploration, and local exploitation as stochastic perturbation and local gradient descent, respectively, to maximize the optimization of segmentation parameters. At the same time, PACF will be adaptively choosing and fusing different classifiers with different feature subsets, which allows the model to react competently to the differences in land structures, atmospheric conditions, and sensor features. To improve the results further we propose a modification called Multi-Objective Gradient-Hopping Hybrid Optimizer (MO-GHHO) + PACF that involves multi-objective optimization to find improved convergence, generalization, and accuracy. The strength and versatility of both models is supported by experiments on high-resolution post-hurricane satellite datasets. The GHHO + PACF model reached the overall accuracy of 95.2% and surpassed the traditional architectures like Visual Geometry Group (VGG)-19, ResNet-50, Inception V3. Also, MO-GHHO + PACF achieves 98.2% of the land classification and 96.1% of the water body segmentation in addition to its significant increase in precision, recall, and F1 score. An ablation study also indicates the effectiveness in the individual contribution of the GHHO and PACF components in the model. The results indicate that the frameworks of GHHO + PACF and MO-GHHO + PACF have a robust, scalable, flexible application to the post-disaster analysis of satellite images and have a great potential as the decision support mechanism to any future rapid damage assessment systems.