Arithmetic Optimization Algorithm Assisted Deep Learning Model for Remote Sensing Image Classification

The spatial resolution of remote sensing images (RSI) was continuously enhanced with the advances in RSI technology. As a fundamental unit of RSI interpretation, the scene is an integration of semantics, multiple objects, and environments. RSI scene understanding not just needs to identify all objects, but needs to observe the topology distribution of many objects in an RSI scene. As a standard representative of deep learning (DL), convolutional neural networks (CNNs) are a promising method to abstract the visual contents of RSI scenes. As a result, CNN was broadly implemented in scene-driven object detection, RSI scene classification, RSI retrieval, etc. This study presents an Arithmetic Optimization Algorithm Assisted Deep Learning Model for Remote Sensing Image Classification (AOADL-RSIC) approach. The presented AOADL-RSIC system exploits Gaussian filtering (GF) approach to eliminate noise. In the presented AOADL-RSIC technique, SqueezeNet model is used for feature extraction process. For the classification of different objects or classes in the RSIs, extreme learning machine (ELM) model is used. Finally, the AOA is applied for the optimum parameter adjustment of the ELM approach. The simulation results of the AOADL-RSIC system are tested on EUROSAT database and the outcome displayed the significance of the AOADL-RSIC approach on image classification process.