Deep Learning-Based Video GIS Behavior Recognition Method Using Cesium

Abstract

The Cesium Video GIS Behavior Recognition Method based on the SlowFast network combines video analysis with Geographic Information Systems (GIS), aiming to enhance behavior recognition in smart city and public safety applications. By utilizing the Cesium platform for video data visualization and spatial data management, and leveraging the dual-path characteristics of the SlowFast model, dynamic behavior features in videos are extracted. The SlowFast model effectively captures video information at different speeds, recognizing behavior patterns with both rapid and slow changes. This method improves the accuracy and robustness of behavior recognition, making it suitable for long-duration video streams and high-resolution data, providing more precise decision support for intelligent monitoring and traffic management.