MUMBAI, India, May 29 -- Intellectual Property India has published a patent application (202631062360 A) filed by Trident Academy Of Technology, Bhubaneswar, Odisha, on May 16, for 'hybrid spectral-attention multi-agent system for pixel-level cloud intelligence with uncertainty quantification.'
Inventor(s) include Kaliprasanna Swain; Sakuntala Mahapatra; Shuvendra Kumar Tripathy; Sumati Baral; Pratiti Mishra; Banaja Mishra; and Dipalika Das.
The application for the patent was published on May 29, under issue no. 22/2026.
According to the abstract released by the Intellectual Property India: "Hybrid Spectral-Attention Multi-Agent System for Pixel-Level Cloud Intelligence with Uncertainty Quantification The present invention discloses a hybrid spectral-attention multi-agent system for reliable pixel-level cloud intelligence in remote sensing imagery. The system employs an SDKBlock-based encoder-decoder backbone with parallel spectral, directional, and multi-scale dilated branches to extract domain-specific features from RGB or multi-spectral satellite data. Four specialized agents Spectral Attention (Transformer), Spatial Context (MobileNet-style), Temporal Sequence (Bidirectional LSTM), and Bayesian Uncertainty operate in parallel on decoder features. Their outputs are integrated through a three-path fusion layer with CBAM attention and dynamically coordinated by a meta-agent orchestrator using per-pixel routing (Softmax) and gating (Sigmoid) mechanisms. The system generates, in a single forward pass, per-pixel probabilities for four classes (clear sky, thick cloud, thin cloud, and shadow) along with aleatoric and epistemic uncertainty maps and rich explainability outputs including routing weights, gating weights, and per-agent logits. This architecture delivers superior accuracy, robustness, and interpretability for cloud detection and classification, addressing critical limitations of existing single-model and generic segmentation approaches in remote sensing applications."
Disclaimer: Curated by HT Syndication.
