Hybrid ResNet152-EML model for Geo-spatial image classification

The performance of categorization for Geo-spatial image on huge data has significantly improved with the development of deep learning (DL) methods. The majority of the Geo-spatial images are complex, with higher intra-class variation and inter-class similarity issues, therefore the performance is still constrained. Numerous approaches are put forth to address these issues; however, some DL algorithms, like Bag of Visual Words (BoVW) and Spatial Pyramid Matching (SPM), require laborious manual feature extraction procedures. Additionally, other DL algorithms require more computational resources for semantic extraction. We suggest a hybrid Residual Network152 model with Extreme Machine Learning (ResNet152-EML) model for end-to-end classification for multi-scale satellite imagery in order to automate the task of feature extraction and improve performance. The original image is broken down into a stack of cropped, Hue Saturation Value (HSV), and Local Binary Pattern (LBP) images by our framework. ResNet152 is used to extract features from these multi-scale sub samples after they have been shrunk into equal dimensions. Extracted features are concatenated and used to train and test the Extreme learning machine. We have compared our model with state-of-art current methods on various datasets. Testing accuracy reported shows the significance of the proposed model is more accurate than any other method. With a hybrid combination, we achieve accuracy up to 99.83, 98.77, 98.54, 98.95, and 98.85% for UC-Merced, RSSCN7, NWPU-RESISC45, WHU-RS19, and AID datasets, respectively. The accuracy achieved is 5 to 10% more than other methods. © The Author(s), under exclusive licence to Bharati Vidyapeeth's Institute of Computer Applications and Management 2023. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.