Probabilistic group-level motion analysis and scenario recognition

Abstract

This paper addresses the challenge of recognizing behavior of groups of individuals in unconstraint surveillance environments. As opposed to approaches that rely on agglomerative or decisive hierarchical clustering techniques, we propose to recognize group interactions without making hard decisions about the underlying group structure. Instead we use a probabilistic grouping strategy evaluated from the pairwise spatial-temporal tracking information. A path-based grouping scheme determines a soft segmentation of groups and produces a weighted connection graph where its edges express the probability of individuals belonging to a group. Without further segmenting this graph, we show how a large number of low-and high-level behavior recognition tasks can be performed. Our work builds on a mature multi-camera multi-target person tracking system that operates in real-time. We derive probabilistic models to analyze individual track motion as well as group interactions. We show that the soft grouping can combine with motion analysis elegantly to robustly detect and predict grouplevel activities. Experimental results demonstrate the efficacy of our approach.

Automatic behavior analysis in tag games: from traditional spaces to interactive playgrounds

“…They cluster people based on the amount of interaction to find relevant groups, and later classify activities. Using a slightly different approach, Chang et al propose using proximity, not levels of interaction, to define groups in their probabilistic model for scenario recognition [11]. They use a soft-grouping approach with path-based connectivity to define group memberships.…”

Group context learning for event recognition

“…Previous works on this type of events mainly use the logic or rule based methods, which require manual creation of rules [3,4,10,14,22] for each event category. Chang et al [4] recognize various group events by combining the results from probabilistic group structure analysis and motion analysis and checking against a list of event models, which are defined manually using scenario-specific predicates. To explicitly model the temporal constraints pertaining to complex events, different probabilistic logical inference engines have been built, such as the Markov Logic Networks [22] and probabilistic event logic [3].…”

“…We then perform group structure analysis of the tracked individuals, as a mean to extract group context features. Following [4], we retain a probabilistic group representation, such that the group-level information can be reliably captured. Specifically, the group analysis produces a weighted connectivity graph for each frame, where the nodes of the graph are the detected individuals and the weight of an edges is the probability of two individuals being in the same group.…”

“…Given the tracking of individuals, we incorporate a probabilistic grouping strategy similar to [4] to perform group analysis and to extract group-level motion and interaction features. As opposed to approaches that rely on hard, agglomerative or divisive clustering techniques to define groups, the probabilistic grouping without a hard segmentation of groups keeps more reliable information about the dynamics of groups that will be later on used to calculate the group context features.…”

“…An overview of our approach is illustrated in Figure 1. We first detect and track people in the video using standard methods [4,30]. Based on the tracking, we analyze the grouplevel structure and motion, and then extract the group context features based on the probabilistic (soft) group structure analysis results ( Figure 3).…”