1,721,667 research outputs found
Hand posture recognition using SURF with adaptive boosting
Full text linkAn approach making use of SURF feature and Adaboost for hand posture recognition is proposed. First the SURF key points are extracted to describe the blob or ridge-like structures from grey level images. These are potential points of interest that can be used to match with other images with similar structures. Then the statistic parameters of the tendency of gradient changes within small patches surrounding the points of interest are calculated as feature vectors. With all the points of interest, Adaboost is used to train a strong classifier for each posture by selecting the most efficient features, which largely lowers the computational cost of the classification stage. The proposed method was tested on the Triesch Hand Posture Database which is the benchmark in the field. Experimental results showed that our method outperforms existing methods in terms of better recognition accuracy
Real-time hand gesture recognition for uncontrolled environments using adaptive SURF tracking and hidden conditional random fields
Full text linkChallenges from the uncontrolled environments are the main difficulties in making hand gesture recognition methods robust in real-world scenarios. In this paper, we propose a real-time and purely vision-based method for hand gesture recognition in uncontrolled environments. A novel tracking method is introduced to track multiple hand candidates from the first frame. The movement directions of all hand candidates are extracted as trajectory features. A modified HCRF model is used to classify gestures. The proposed method can survive challenges including: gesturing hand out of the scene, pause during gestures, complex background, skin-coloured regions moving in background, performers wearing short sleeve and face overlapping with hand. The method has been tested on Palm Graffiti Digits database and Warwick Hand Gesture database. Experimental results show that the proposed method can perform well in uncontrolled environments
A framework for real-time hand gesture recognition in uncontrolled environments with partition matrix model based on hidden conditional random fields
Full text linkThe main obstructions of making hand gesture recognition methods robust in real-world applications are the challenges from the uncontrolled environments, including: gesturing hand out of the scene, pause during gestures, complex background, skin-coloured regions moving in background, performers wearing short sleeve and face overlapping with hand. Therefore, a framework for real-time hand gesture recognition in uncontrolled environments is proposed in this paper. A novel tracking scheme is proposed to track multiple hand candidates in unconstrained background, and a weighting model for gesture classification based on Hidden Conditional Random Fields which takes trajectories of multiple hand candidates under different frame rates into consideration is also introduced. The framework achieved invariance under change of scale, speed and location of the hand gestures. The Experimental results of the proposed framework on Palm Graffiti Digits database and Warwick Hand Gesture database show that it can perform well in uncontrolled environments
Located Lexicon: a project that explores how user generated content describes place
No full textThis extended conference paper explores the use and potential of location data in social media contexts. The research involved a series of experiments undertaken to assess the extent to which location information is present in exchanges, directly or indirectly. A prototype application was designed to exploit the insight obtained from the data-gathering experiments. This enabled us to develop a method and toolkit for searching, extracting and visualising mass-generated data for open source use. Ultimately, we were able to generate insights into data quality and ‘scale of query’ for emerging pedagogical research in learning swarms and distributed learners
3D Reconstruction of freely moving persons for re-identification with a depth sensor
No full textIn this work, we describe a novel method for
creating 3D models of persons freely moving in front of a
consumer depth sensor and we show how they can be used
for long-term person re-identification. For overcoming the
problem of the different poses a person can assume, we exploit
the information provided by skeletal tracking algorithms for
warping every point cloud frame to a standard pose in real
time. Then, the warped point clouds are merged together to
compose the model. Re-identification is performed by matching
body shapes in terms of whole point clouds warped to a
standard pose with the described method. We compare this
technique with a classification method based on a descriptor
of skeleton features and with a mixed approach which exploits
both skeleton and shape features. We report experiments on
two datasets we acquired for RGB-D re-identification which
use different skeletal tracking algorithms and which are made
publicly available to foster research in this new research branch
Semantic concept detection in imbalanced datasets based on different under-sampling strategies
Full text linkSemantic concept detection is a very useful technique for developing powerful retrieval or filtering systems for multimedia data. To date, the methods for concept detection have been converging on generic classification schemes. However, there is often imbalanced dataset or rare class problems in classification algorithms, which deteriorate the performance of many classifiers. In this paper, we adopt three “under-sampling” strategies to handle this imbalanced dataset issue in a SVM classification framework and evaluate their performances
on the TRECVid 2007 dataset and additional positive
samples from TRECVid 2010 development set. Experimental
results show that our well-designed “under-sampling” methods
(method SAK) increase the performance of concept detection
about 9.6% overall. In cases of extreme imbalance in
the collection the proposed methods worsen the performance
than a baseline sampling method (method SI), however in the
majority of cases, our proposed methods increase the performance of concept detection substantially. We also conclude that method SAK is a promising solution to address the SVM classification with not extremely imbalanced datasets
Goal accomplishment tracking for automatic supervision of plan execution
Full text linkIt is common practice to break down plans into a series of goals or sub-goals in order to facilitate plan execution, thereby only burdening the individual agents responsible for their execution with small, easily achievable objectives at any one time, or providing a simple way of sharing these objectives amongst a group of these agents. Ensuring that plans are executed correctly is an essential part of any team management. To allow proper tracking of an agent's progress through a pre-planned set of goals, it is imperative to keep track of which of these goals have already been accomplished. This centralised approach is essential when the agent is part of a team of humans and/or robots, and goal accomplishment is not always being tracked at a low level. This paper presents a framework for an automated supervision system to keep track of changes in world states so as to chart progress through a pre-planned set of goals. An implementation of this framework on a mobile service robot is presented, and applied in an experiment which demonstrates its feasibility
Algorithms for people re-identification from RGB-D videos exploiting skeletal information
Full text linkIn this thesis a novel methodology to face people re-identification problem is proposed. Re-identification is a complex research topic representing a fundamental issue especially for intelligent video surveillance applications. Its goal is to determine the occurrences of the same person in different video sequences or images, usually by choosing from a high number of candidates within a datase
Combining traffic sign detection with 3D tracking towards better driver assistance
No full textWe briefly review the advances in driver assistance systems and present a real-time version that integrates single view detection with region-based 3D tracking of traffic signs. The system has a typical pipeline: detection and recognition of traffic signs in independent frames, followed by tracking for temporal integration. The detection process finds an optimal set of candidates and is accelerated using AdaBoost cascades. A hierarchy of SVMs handles the recognition of traffic sign types. The 2D detections are then employed in simultaneous 2D segmentation and 3D pose tracking, using the known 3D model of the recognized traffic sign. Thus, we achieve not only 2D tracking of the recognized traffic signs, but we also obtain 3D pose information, which we use to establish the relevance of the traffic sign to the driver. The performance of the system is demonstrated by tracking multiple road signs in real-world scenarios.Radu Timofte, Victor Adrian Prisacariu, Luc Van Gool, and Ian Rei
Interoperable multimedia mobile services for cultural heritage sites
Full text linkAccording to the ancient Romans, “Delectare, docere, movere” are the goals of eloquence. To be accepted by museums, landscapes and archaeological sites, technology has to win the same challenge. Is technology unobtrusive enough to avoid compromising the emotional involvement that makes a visit to a cultural site unforgettable? Can it achieve a dissemination of the information in
such a way that it is understood better? And how can technology be used to increase visibility and understanding of the numerous sites that are not yet able to attract the amount of people they deserve?
This paper presents the authors’ vision on these questions, reporting on the activities carried out by the “mobile and ambient systems” work group of EPOCH as part of the CIMAD project.
A central part of CIMAD is the creation of services for visitors and archaeological sites as well as making parts of the overall
vision a reality. The CIMAD services are based around the MobiComp context infrastructure, enabling the services to exchange context information and information to be displayed to the user. As the EPOCH network is beginning to dissolve we will discuss possible next steps, associated risks and opportunities of continuing this project
- …
