MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641095330 A) filed by Vardhaman College Of Engineering on August 06, 2026, for Autonomous Satellite Image Interpretation Using Self-Supervised Vision Transformers.
Inventors include Dr. Sekkeri Venu Gopal; Mr. V N S Vijaya Kumar; Ms. Chalamani Bhavana; Ms. Baalne Anjali; Mr. Kadulla Haribabu; and Ms. Asma Begum.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: Autonomous Satellite Image Interpretation Using Self-Supervised Vision Transformers is the proposed invention. The invention provides an autonomous satellite image interpretation system that utilises a Self-Supervised Vision Transformer (SS-ViT) to intelligently analyse remote sensing images without the need for large manually labelled datasets. The system acquires multispectral, hyperspectral, optical, thermal, and synthetic aperture radar images, carries out preprocessing and image patch generation, and uses masked image modelling with contrastive self-supervised learning to extract robust spatial and spectral features. The pretrained Vision Transformer is fine-tuned for land-use classification, semantic segmentation, object detection, environmental monitoring, disaster assessment, infrastructure mapping, and change detection. The multi-head self-attention mechanism can model long-range contextual relationships in satellite scenes accurately, which can boost the interpretation accuracy compared with conventional convolution-based methods. An uncertainty estimation module identifies predictions of low confidence and an Explainable Artificial Intelligence module generates attention maps that highlight the parts of the image that influence the decisions of the model.
Disclaimer: Curated by HT Syndication.
