Probabilistic Latent Sequential Motifs: Discovering Temporal Activity Patterns in Video Scenes.

Abstract

This paper introduces a novel probabilistic activity modeling approach that mines recurrent sequential patterns from documents given as word-time occurrences. In this model, documents are represented as a mixture of sequential activity motifs (or topics) and their starting occurrences. The novelties are threefold. First, unlike previous approaches where topics only modeled the co-occurrence of words at a given time instant, our topics model the co-occurrence and temporal order in which the words occur within a temporal window. Second, our model accounts for the important case where activities occur concurrently in the document. And third, our method explicitly models with latent variables the starting time of the activities within the documents, enabling to implicitly align the occurrences of the same pattern during the joint inference of the temporal topics and their starting times. The model and its robustness to the presence of noise have been validated on synthetic data. Its effectiveness is also illustrated in video activity analysis from low-level motion features, where the discovered topics capture frequent patterns that implicitly represent typical trajectories of scene objects.

Multi-camera open space human activity discovery for anomaly detection

“…This class of approaches have the ability to discover dominant activity patterns occurring in the scene using simple low-level features. In this article, we use one such model called Probabilistic Latent Sequential Motifs (PLSM) [9]. PLSM, unlike other topic models, represents activities as temporal patterns called motifs.…”

“…Existing topic models were demonstrated only on highly constrained scenes such as traffic scenes and with a single viewpoint [5,9,4]. In our case, we explore the use of these models in the context of a metro station that contains multiple cameras and loosely constrained activities.…”

“…An abnormality can be roughly categorized into two types [7]: 1) events that are fundamentally unusual in appearance, and 2) unusual order of events, where all or most of these events are normal. The adopted PLSM model [9], due to its complete representation of scene activities, can simultaneously capture both kinds of abnormalities.…”

“…Compared to [9], we use different low level features and a different low-level processing. To avoid the need for a complex background model, we only use quantized optical flow features.…”

A combination of generative and discriminative models for fast unsupervised activity recognition from traffic scene videos

“…To overcome this problem and to demonstrate the capability of our fast HDP techniques, we use the Traffic dataset of [17]. This video contains over 44 minutes of recording of road traffic with a resolution of 360 × 288 pixels per frame.…”