MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202621073274 A) filed by Prashant Sheshrao Titare; Prabhu Reddy; Sarthak Manish Ambhore; Anush Sachin Bedkyale; and Sajid Sayyed on June 12, 2026, for Ai-Powered Visual Detection Of Structural Defects In Industrial Warehouse Racking Systems Using Deep Learning.
Inventors include Prashant Sheshrao Titare; Prabhu Reddy; Sarthak Manish Ambhore; Anush Sachin Bedkyale; and Sajid Sayyed.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: An automated visual inspection system and method for the real-time detection, localization, and classification of structural anomalies in industrial warehouse racking systems. The invention utilizes a single-stage, anchor-free convolutional neural network (CNN) to automatically identify critical safety hazards, such as dents, bends, rust, and severe structural deformation, from optical data. To maintain high predictive accuracy in challenging warehouse environments, the system incorporates an adaptive preprocessing engine that applies Contrast Limited Adaptive Histogram Equalization (CLAHE) to the luminance channel and utilizes aspect-preserving letterbox resizing to prevent geometric distortion. Furthermore, the neural network employs a tri-fold imbalance resolution architecture comprising asymmetric data augmentation, inverse-frequency class weighting, and dynamic focal loss tuning to guarantee precise detection of rare, catastrophic structural failures. The computational graph is exported via an optimized runtime environment, enabling ultra-low latency inference suitable for integration into continuously moving mobile inspection platforms, such as drones and Automated Guided Vehicles (AGVs).
Disclaimer: Curated by HT Syndication.
