Catálogo de publicaciones - libros

This paper proposes a system for tracking people in video streams, returning their body and head bounding boxes. The proposed system comprises a variation of Stauffer’s adaptive background algorithm with spacio-temporal adaptation of the learning parameters and a Kalman tracker in a feedback configuration. In the feed-forward path, the adaptive background module provides target evidence to the Kalman tracker. In the feedback path, the Kalman tracker adapts the learning parameters of the adaptive background module. The proposed feedback architecture is suitable for indoors and outdoors scenes with varying background and overcomes the problem of stationary targets fading into the background, commonly found in variations of Stauffer’s adaptive background algorithm.

This paper describes Pittsburgh Pattern Recognition’s participation in the face detection and tracking tasks for the CLEAR 2006 evaluation. We first give a system overview, briefly explaining the three main stages of processing: (1) frame-based face detection; (2) motion-based tracking; and (3) track filtering. Second, we summarize and analyze our system’s performance on two test data sets: (1) the CHIL Interactive Seminar corpus, and (2) the VACE Multi-site Conference Meeting corpus. We note that our system is identically configured for all experiments, and, as such, makes use of no site-specific or domain-specific information; only video continuity is assumed. Finally, we offer some concluding thoughts on future evaluations.

This paper presents the tracking system from Athens Information Technology that participated to the pedestrian and vehicle surveillance task of the CLEAR 2006 evaluations. Two are the novelties of the proposed tracker. First, we use a variation of Stauffer’s adaptive background algorithm with spatiotemporal adaptation of the learning parameters and a Kalman filter in a feedback configuration. In the feed-forward path, the adaptive background module provides target evidence to the Kalman filter. In the feedback path, the Kalman filter adapts the learning parameters of the adaptive background module. Second, we combine a temporal persistence pixel map, together with edge information, to produce the evidence that is associated with targets. The proposed tracker performed well in the evaluations, and can be also applied to indoors settings and multi-camera tracking.

The evaluation results of a system for tracking humans in surveillance videos are presented. Moving blobs are detected based on adaptive background modeling. A shape based multi-view human detection system is used to find humans in moving regions. The detected responses are associated to infer the human trajectories. The shaped based human detection and tracking is further enhanced by a blob tracker to boost the performance on persons at a long distance from the camera. Finally the 2D trajectories are projected onto the 3D ground plane and their 3D speeds are used to verified the hypotheses. Results are given on the video test set of the VACE surveillance human tracking evaluation task.

We present an object detection and tracking algorithm that addresses the problem of multiple simultaneous targets tracking in real-world surveillance scenarios. The algorithm is based on color change detection and multi-feature graph matching. The change detector uses statistical information from each color channel to discriminate between foreground and background. Changes of global illumination, dark scenes, and cast shadows are dealt with a pre-processing and post-processing stage. Graph theory is used to find the best object paths across multiple frames using a set of weighted object features, namely color, position, direction and size. The effectiveness of the proposed algorithm and the improvements in accuracy and precision introduced by the use of multiple features are evaluated on the VACE dataset.

In this paper, we present KNIGHT, a Windows-based stand-alone object detection, tracking and classification software, which is built upon Microsoft Windows technologies. The object detection component assumes stationary background settings and models background pixel values using Mixture of Gaussians. Gradient-based background subtraction is used to handle scenarios of sudden illumination change. Connected- component algorithm is applied to detected foreground pixels for finding object-level moving blobs. The foreground objects are further tracked based on a pixel-voting technique with the occlusion and entry/exit reasonings. Motion correspondences are established using the color, size, spatial and motion information of objects. We have proposed a texture-based descriptor to classify moving objects into two groups: vehicles and persons. In this component, feature descriptors are computed from image patches, which are partitioned by concentric squares. SVM is used to build the object classifier. The system has been used in the VACE-CLEAR evaluation forum for the vehicle tracking task. Corresponding system performance is presented in this paper.

This paper describes a system for tracking people and vehicles for stationary-camera visual surveillance. The appearance of objects being tracked is modeled using mixtures of mixtures of Gaussians. Particles filters are used to track the states of object. Results show the robustness of the system to various lighting and object conditions.

Evaluation results of a vehicle tracking system on a given set of evaluation videos of a street surveillance system are presented. The method largely depends on detection of motion by comparison with a learned background model. Several difficulties of the task are overcome by the use of general constrains of scene, camera and vehicle models. An analysis of results is also presented.

In this paper the person identification system developed at Athens Information Technology is presented. It comprises of an audio-only (speech), a video-only (face) and an audiovisual fusion subsystem. Audio recognition is based on the Gaussian Mixture modeling of the principal components of the Mel-Frequency Cepstral Coefficients of speech. Video recognition is based on linear subspace projection methods and temporal fusion using weighted voting on the results. Audiovisual fusion is done by fusing the unimodal identities into the multimodal one, using a suitable confidence metric for the results of the unimodal classifiers.

This paper summarizes the LIMSI participation in the CLEAR’06 acoustic speaker identification task that aims to identify speakers in CHIL seminars via the acoustic channel. The system consists of a standard Gaussian mixture model based system similar to systems developed for the NIST speaker recognition evaluations and includies feature warping of cepstral coefficients and MAP adaptation of a Universal Background Model. Several computational optimizations were implemented for real-time efficiency: stochastic frame subsampling for training, top-Gaussians scoring and auto-adaptive pruning for the tests, speeding up the system by more than a factor of ten.