BieColl - Bielefeld eCollections
Not a member yet
1004 research outputs found
Sort by
An Un-awarely Collected Real World Face Database: The LabName-Door Face Database
No full textIn this paper we present a new face database that has been collected under real world conditions and without collaborating with the individuals whose images are being captured. The images in the database are recorded with a zoom camera monitoring the door of the laboratory. The developed capture software processes each frame and whenever it detects a face as well as the eyes, it saves the frame to the database. A face recognition software is also accompanied to the image acquisition system to help to label the identities of the individuals in the database. Recordings have been done for six months and this way ten thousands pictures of more than 100 individuals have been collected. From this set, approximately 33000 images of 30 people are made available to public. To give an idea about the difficulty level of doing face recognition under such a scenario, the well-known face recognition algorithms are tested on the collected database
Automatic Calibration of a Urban Video Surveillance System through the Observation of Zebra Crossings
Get PDFIn this paper, a method for automatic calibration of a camera stereo pair through the observation of zebra crossing signs is described. It is based on the well-known consideration that it is possible to obtain information about lens distortion and camera orientation by observing how a known pattern appears in the image; moreover, a major advantage of this system is that it does not require any ad-hoc calibration pattern, because it exploits the zebra crossing signs, a pattern usually present in images used for monitoring pedestrians while crossing a road. To achieve this goal, well-known techniques for removing lens distortion and perspective effect are combined with new methods for locating calibration points on the available pattern, and, finally, for evaluating the camera position
Classifier training based on synthetically generated samples
Get PDFIn most image classification systems, the amount and quality of the training samples used to represent the different pattern classes are important factors governing the recognition performance. Hence, it is usually necessary to acquire a representative set of training samples by acquisition of data in real-world environments. Such procedures may require considerable efforts and furthermore often generate a training set which is unbalanced with respect to the number of available samples per class. In this contribution we regard classification tasks for which each real-world training sample is derived from an ideal class representative which undergoes a geometric and photometric transformation. This transformation depends on system-specific influencing quantities of the image formation process such as illumination, characteristics of the sensor and optical system, or camera motion. The parameters of the transformation model are learned from object classes for which a large number of real-world samples are available. For each individual real-world sample a set of model parameters is derived by correspondingly fitting the transformed ideal sample to the observed sample. The obtained probability distribution of model parameters is used to generate synthetic sample sets for all regarded pattern classes. This training approach is applied to a vehicle-based vision system for traffic sign recognition. Our experimental evaluation on a large set of real-world test data demonstrates that the classification rates obtained for classifiers trained with synthetic samples are comparable to those obtained based on real-world training data
Gait-Based Pedestrian Detection for Automated Surveillance
Get PDFIn this paper, we explore a new approach for walking pedestrian detection in an unconstrained outdoor environment. The proposed algorithm is based on gait motion as the rhythm of the footprint pattern of walking people is considered the stable and characteristic feature for the classification of moving objects. The novelty of our approach is motivated by the latest research for people identification using gait. The experimental results confirmed the robustness of our method to discriminate between single walking subject, groups of people and vehicles with a detection rate of %100. Furthermore, the results revealed the potential of our method to extend visual surveillance systems to recognize walking people
GPAPF: A Combined Approach for 3D Body Part Tracking
Get PDFIn this paper we present a combined approach for body part tracking in 3D using multiple cameras, called GPAPF. This approach combines annealed particle filtering, which has been shown as effective tracker for body parts, with Gaussian Process Dynamical Model, which is used in order to reduce the dimensionality of the problem. That reduction improves the tracker's performance and increases the tracker's stability and ability to recover from the loosing the target. We also compare GPAPF tracker with the annealed particle filter and show that our tracker has a better performance even for low frame rate sequences
On-line Learning-based Object Tracking Using Boosted Features
Get PDFThe most informative and hard to classify examples are close to the decision boundary between object of interest and background. Gentle AdaBoost built on regression stumps focuses on hard examples that provide most new information during object tracking. They contribute to better learning of the classifier while tracking the object. The tracker is compared to recently proposed algorithm that uses on-line appearance models. The performance of the algorithm is demonstrated on freely available test sequences. The resulting algorithm runs in real-time
Simultaneously Reconstructing Transparent and Opaque Surfaces from Texture Images
Get PDFThis paper addresses the problem of reconstructing non-overlapping transparent and opaque surfaces from multiple view images. The reconstruction is attained through progressive refinement of an initial 3D shape by minimizing the error between the images of the object and the initial 3D shape. The challenge is to simultaneously reconstruct both the transparent and opaque surfaces given only a limited number of images. Any refinement methods can theoretically be applied if analytic relation between pixel value in the training images and vertices position of the initial 3D shape is known. This paper investigates such analytic relations for reconstructing opaque and transparent surfaces. The analytic relation for opaque surface follows diffuse reflection model, whereas for transparent surface follows ray tracing model. However, both relations can be converged for reconstruction both surfaces into texture mapping model. To improve the reconstruction results several strategies including regularization, hierarchical learning, and simulated annealing are investigated
Spatio-Temporal Reasoning for Reliable Facial Expression Interpretation
Get PDFUnderstanding human behaviours and emotions has received contributions from image analysis and pattern recognition techniques in order to tackle this challenge. The most popular facial expression classifiers deal with eyebrows and lips while avoiding eyelid motion. According to psychologists, eye motion is relevant for trust and deceit analysis as well for dichotomizing near facial expressions. Unlike previous approaches, we include the eyelid motion by constructing an appearance-based tracker (ABT). Subsequently, a Case-Based Reasoning (CBR) approach is applied by training a case-base with seven facial actions. We classify new facial expressions with respect to previous solutions, previously assessing confidence for the proposed solutions. Therefore, the proposed system yields efficient classification rates comparable to the best previous facial expression classifiers. The ABT and CBR combination provides trusty solutions by evaluating the confidence of the solution quality for eyebrows, mouth and eyes. Consequently, this method is robust and accurate for facial motion coding, and for confident classifications. The training is progressive, the quality of the solution increases with respect to previous solutions and do not need re-training processes
Task and Context aware Performance Evaluation of Computer Vision Algorithms
Get PDFDeveloping a robust computer vision algorithm is very difficult because of the enormous variation of visual conditions. A systems technology solution to this challenge is an automatic selection and configuration of different existing algorithms according to the task and context of arbitrary applications. This paper presents a first attempt to generate the required mapping between the task/context to the optimal algorithm and algorithm configuration. This mapping is based on an extensive performance evaluation. To practically handle the exhaustive search for optimal solutions a new optimization challenge the Multiple-Multi Objective Optimization (M-MOP) and an according solution based on genetic algorithms is developed and evaluated. The results show the robustness of the approach and guide further development towards an automatic vision system generation
A news-based financial time series discretization
Get PDFIn this paper a new method for financial time series discretization that allows to take into account qualitative features about financial indicators is proposed. Qualitative features are extracted from financial news web sites and they are inserted into the learning phase of a recursive Self Organizing Map by means of a suitable parameter derived from statistical analysis of document collections. A postprocessing phase based on unsupervised clustering by U-Matrix method leads to the actual discretization of the time series. A real case application to a stock closing price series reveals that the inclusion of qualitative features leads to a more compact discretization of the series. This could be useful if a compact coding of the series is sought, for example in the preprocessing phase of a forecasting methodology