MUMBAI, India, May 29 -- Intellectual Property India has published a patent application (202641064008 A) filed by Vellore Institute Of Technology, Vellore, Tamil Nadu, on May 21, for 'a multiclass video anomaly detection method and a system for intelligent surveillance.'
Inventor(s) include Bhawana Tyagi; Suvandita Swaroop; and Kajal Pahil.
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: "The present invention relates to a multiclass video anomaly detection system and method for intelligent surveillance are disclosed. the system includes at least one video capture unit for acquiring continuous video streams from fixed surveillance cameras and a processing unit configured to perform frame-wise analysis of the video streams. a YOLOv7-based spatial feature extraction module detects objects in each frame and generates high-dimensional spatial embeddings including object appearance features, bounding-box coordinates, and confidence values, wherein the YOLOv7 module functions solely as a feature encoder. a temporal sequence formation module arranges the frame-level embeddings into fixed-length, chronologically ordered temporal windows representing behavioral sequences. a spatiotemporal learning module comprises a convolutional neural network (CNN) branch for learning spatial interaction cues and a long short-term memory (LSTM) branch for learning temporal evolution patterns. outputs of the CNN and LSTM branches are combined by a feature fusion module to form a unified latent representation, which is processed by a multiclass anomaly classification module to generate probabilities for multiple anomaly categories. the system and method enable real-time and offline detection and classification of multiple distinct anomaly types, including vandalism, shoplifting, and arson, from surveillance video streams."
Disclaimer: Curated by HT Syndication.
