Remote sensing scene classification using visual geometry group 19 model and multi objective grasshopper optimization algorithm

Recently, the Remote Sensing Scene Classification (RSSC) has played a vital role in several applications: environment monitoring, urban planning, and land management. The deep neural networks are extensively utilized in the RSSC, because of their superior performance. In recent decades, several scene classification models improve classification accuracy by incorporating extra modules, but it increases the computing overhead and parameters of the models at the inference phase. In addition, the complementarity of the features extracted by the deep learning models is exploited to reduce the improvement of classification accuracy. For addressing the aforementioned issues, a new meta-heuristics based Visual Geometry Group-19 (VGG-19) model is implemented in this research manuscript. After acquiring the aerial images from REmote Sensing Image Scene Classification 45 (RESISC45), Aerial Image Dataset (AID) and the University of California Merced (UC Merced) datasets, the VGG-19 network is applied for classifying the scene categories. In the proposed system, a multi-objective Grasshopper Optimization Algorithm (GOA) is implemented for selecting the optimal hyper-parameters of the VGG-19 model, which helps in reducing the computational complexity and training time of the model. The experimental results demonstrated that the meta-heuristics based VGG-19 model achieved 98.67%, 99.57%, and 98.06% of accuracy on the AID, UC Merced, and RESISC45 datasets, which are superior related to the comparative deep learning models.