Convolution Network for Relevant Motion Detection in Surveillance Video

The USPTO published patent application US20260099927A1 covering methods and systems for detecting relevant motion of objects of interest (persons and vehicles) in surveillance videos. The invention uses input data from captured images/video that undergoes pre-processing before being fed into a convolution network with both spatial-wise and temporal-wise max pooling capabilities to generate prediction results.

For affected parties, this patent application represents a potential intellectual property filing in the AI/surveillance technology space. Technology companies developing video analytics, surveillance systems, or computer vision applications should monitor this application for potential overlap with their own R&D activities. The patent does not create immediate compliance obligations but may inform future patent strategy or competitive intelligence in AI-based motion detection technology.