Research on Identifying Differential Behavior in Complex Scenarios

Group behavior recognition is a significant research area in computer vision and pattern recognition. However, recognizing an individual who behaves differently from the majority in complex scenes, such as malls and transport stations, is even more crucial. This paper proposes a novel approach for group behavior recognition based on deep learning and graph theory. The approach comprises two main stages: feature extraction and graph-based classification. In the feature extraction stage, visual features are extracted from each individual in the group. In the graph-based classification stage, a graph representation of the group is constructed, and graph convolutional networks (GCNs) are utilized to classify the group behavior and differentiate the anomalous behavior.