BieColl - Bielefeld eCollections
Not a member yet
1004 research outputs found
Sort by
Failure detection and separation in SOM based decision support
Get PDFFailure management in process industry has difficult tasks. Decision support in control rooms of nuclear power plants is needed. A prototype that uses Self-Organizing Map (SOM) method is under development in an industrial project. This paper has focus on failure detection and separation. A literature survey outlines the state-of-the-art and reflects our study to related works. Different SOM visualizations are used. Failure management scenarios are carried out to experiment the methodology and the Man-Machine Interface (MMI). U-matrix trajectory analysis and quantization error are discussed more in detail. The experiments show the usefulness of the chosen approach. Next step will be to add more practical views by analyzing real and simulated industrial data with the control room tool and by feedback from the end users
Single pass clustering for large data sets
Get PDFThe presence of very large data sets poses new problems to standard neural clustering and visualization algorithms such as Neural Gas (NG) and the Self-Organizing-Map (SOM) due to memory and time constraints. In such situations, it is no longer possible to store all data points in the main memory at once and only a few, ideally only one run over the whole data set is still affordable to achieve a feasible training time. In this contribution we propose single pass extensions of the classical clustering algorithms NG and fuzzy-k-means which are based on a simple patch decomposition of the data set and fast batch optimization schemes of the respective cost function. The algorithms maintain the benefits of the original ones including easy implementation and interpretation as well as large flexibility and adaptability because of the underlying cost function. We demonstrate the efficiency of the approach in a variety of experiments
Task Segmentation in a Mobile Robot by mnSOM and Hierarchical Clustering
Get PDFOur previous studies assigned labels to mnSOM modules based on the assumption that winner modules corresponding to subsequences in the same class share the same label. We propose segmentation using hierarchical clustering based on the resulting mnSOM. Since it does not need the above unrealistic assumption, it gains practical importance at the sacrifice of the deterioration of the segmentation performance by 1.2%. We compare the performance of task segmentation for two kinds of module architecture in mnSOM. The result is that module architecture with sensory-motor signals as target outputs has superior performance to that with only sensory signals as target outputs
A Layered Active Memory Architecture for Cognitive Vision Systems
No full textRecognising actions and objects from video material has attracted growing research attention and given rise to important applications. However, injecting cognitive capabilities into computer vision systems requires an architecture more elaborate than the traditional signal processing paradigm for information processing. Inspired by biological cognitive systems, we present a memory architecture enabling cognitive processes (such as selecting the processes required for scene understanding, layered storage of data for context discovery, and forgetting redundant data) to take place within a computer vision system. This architecture has been tested by automatically inferring the score of a tennis match, and experimental results show a significant improvement in the overall vision system performance --- demonstrating that managing visual data in a manner more akin to that of the human brain is a key factor in improving the efficiency of computer vision systems
Digitalisation of Warped Documents Supported by 3D-Surface Reconstruction
Get PDFThe high quality digitalisation of warped documents is still a big problem for most scanner technologies. The presented work is a contribution to develop a new technique handling this problem. Basic principle of the proposed method is a special kind of light section, that works with a comparatively very broad stripe lighting by using one additional matrix camera. We can reconstruct the 3d-surface of the document by simple capturing an image sequence of the stripe lighting of a common book scanner during the scanning process. Based on a surface model we transform the warped document in a plane. Result is the two-dimensional output being a nearly distortion-free digital copy of the original warped document
Exploratory Learning Strucutre in Artificial Cognitive Systems
Get PDFOne major goal of the COSPAL project is to develop an artificial cognitive system architecture with the capability of exploratory learning. Exploratory learning is a strategy that allows to apply generalization on a conceptual level, resulting in an extension of competences. Whereas classical learning methods aim at best possible generalization, i.e., concluding from a number of samples of a problem class to the problem class itself, exploration aims at applying acquired competences to a new problem class. Incremental or online learning is an inherent requirement to perform exploratory learning. Exploratory learning requires new theoretic tools and new algorithms. In the COSPAL project, we mainly investigate reinforcement-type learning methods for exploratory learning and in this paper we focus on its algorithmic aspect. Learning is performed in terms of four nested loops, where the outermost loop reflects the user-reinforcement-feedback loop, the intermediate two loops switch between different solution modes at symbolic respectively sub-symbolic level, and the innermost loop performs the acquired competences in terms of perception-action cycles. We present a system diagram which explains this process in more detail. We discuss the learning strategy in terms of learning scenarios provided by the user. This interaction between user ('teacher') and system is a major difference to many existing systems where the system designer places his world model into the system. We believe that this is the key to extendable robust system behavior and successful interaction of humans and artificial cognitive systems. We furthermore address the issue of bootstrapping the system, and, in particular, the visual recognition module. We give some more in-depth details about our recognition method and how feedback from higher levels is implemented. The described system is however work in progress and no final results are available yet. The available preliminary results that we have achieved so far, clearly point towards a successful proof of the architecture concept
Integrating Face-ID into an Interactive Person-ID Learning System
Get PDFAcquiring knowledge about persons is a key functionality for humanoid robots. By envisioning a robot that can provide personalized services the system needs to detect, recognize and memorize information about specific persons. To reach this goal we present an approach for extensible person identification based on visual processing, as one component of an interactive system able to interactively acquire information about persons. This paper describes an approach for face-ID recognition and identification over image sequences and its integration into the interactive system. We compare the approach of sequence hypotheses against results from single image hypotheses, and a standard approach and show improvements in both cases. We furthermore explore the usage of confidence scores to allow other system components to estimate the accuracy of face-ID hypotheses
Navigation of Nonholonomic Mobile Robot Using Visual Potential Field
Get PDFIn this paper, we develop an algorithm for the navigation of a nonholonomic mobile robot using the visual potential. The robot is equipped with a camera system which dynamically captures the environment. The visual potential is computed from an image sequence and optical flow computed from successive images captured by the camera mounted on the robot. Our robot selects a local pathway using the visual potential computed from its vision system without any knowledge of a robot workspace. We present experimental results of the obstacle avoidance in the real environment
Radiometric alignment and vignetting calibration
Get PDFThis paper describes a method to photometrically align registered and overlapping images which have been subject to vignetting (radial light falloff), exposure variations, white balance variation and nonlinear camera response. Applications include estimation of vignetting and camera response; vignetting and exposure compensation for image image mosaicing; and creation of high dynamic range mosaics. Compared to previous work white balance changes can be compensated and a computationally efficient algorithm is presented. The method is evaluated with synthetic and real images and is shown to produce better results than comparable methods
Real-time scattering compensation for time-of-flight camera
Get PDF3D images from time-of-flight cameras may suffer from false depth readings caused by light scattering. In order to reduce such scattering artifacts, a scattering compensation procedure is proposed. First, scattering is analysed and expressed as a linear transform of a complex image. Then, a simple scattering model is formulated. Assuming a space invariant point spread function as a model for the scattering leads to a solution in a form of a deconvolution scheme whose computational feasibility and practical applicability are further discussed in this paper