Claus Neubauer

License plate recognition with an intelligent camera

This paper describes a robust car identification system for surveillance of parking lot entrances that runs completely on a stand alone low cost intelligent camera. In order to meet accuracy and speed requirements hierarchical classifiers and coarse to fine search techniques are applied at each recognition stage from localization, segmentation to classification. The paper gives an overview of the applied image processing techniques and focuses in particular on the character classification part. Character recognition is based on a convolutional neural network that proved to generalize better than a fully connected multi-layer perceptron.

Support vector methods and use of hidden variables for power plant monitoring

This paper has three contributions to the fields of power plant monitoring. First, we differentiate out-of-range detection from fault detection. An out-of-range refers to a normal operating range of a power plant unseen in the training data. In the case of an out-of-range, instead of producing a fault alarm, the system should notify the operator to include more training data which capture this new operating range. Second, we apply a support vector one-class classifier to out-of-range detection for its good volume modeling ability. Third, we propose to use hidden variables in regression models for fault detection. This is shown to be much better than prior work in terms of spillover reduction.

Development of Data Infrastructure for the Long Term Bridge Performance Program

The growth of the National Bridge Inventory database, the availability of new data from embedded sensors, in-situ tests and live load tests together with additional bridge related data, e.g. geospatial and weather data, represents an immense source of information helpful for a better understanding of bridge performance and deterioration. However, in order to efficiently exploit this overwhelming amount of information a new generation of data management and data analysis tools is needed. In this paper we describe an open, scalable, and extensible data management and data analysis infrastructure which will be established within the framework of the Long Term Bridge Performance Program (LTBP). MOTIVATION Highway bridges play an important role in the national transportation network. Like any other infrastructure asset, bridges deteriorate with time and require regular maintenance to continue operating at an acceptable level. Often, funding constraints makes it impossible for the bridge owners to perform all maintenance activities that should be carried out at a given time, and they must face the difficult decision of selecting a small subset of maintenance activities that could be performed within available resources, while maximizing the return on investment. Making educated decisions on maintenance activities require bridge owners and other stakeholders to better understand bridge performance. More specifically, given the current condition of a bridge, owners must understand how each of the recommended maintenance activities may impact the overall bridge function, and which activities provide the best costbenefit tradeoff. However, understanding bridge performance is non-trivial and requires an indepth analysis of how bridges function and behave under various complex and interrelated factors and stresses, including but not limited to traffic volumes, overall load, weather conditions, environmental assaults, age, materials, design and prior maintenance history of the bridge.

A Variational Bayesian Approach for Classification with Corrupted Inputs

Classification of corrupted images, for example due to occlusion or noise, is a challenging problem. Most existing methods tackled this problem using a two-step strategy: image reconstruction and classification of reconstructed images. However, their performances heavily relied on the accuracy of reconstruction and parameter estimation. We present a full Bayesian approach which infers the class label from the corrupted image by marginalizing the original image and parameters. Overfitting is effectively overcome through Bayesian integration. Our system consists of two models. The original image model, which specifies the original image generation process, is described by a Gaussian mixture model. The observation model, which relates the corrupted image to the original image, is depicted by an additive deviation model. Normal pixel and corrupted pixel values are elegantly handled by the covariance of the Gaussian deviation. We employ variational approximation to make the Bayesian integration tractable. The advantage of the proposed method is demonstrated by classification tests on the USPS digit database and PIE face database with pose and illumination variations.

Robust sensor estimation using temporal information

We propose a dynamic Bayesian framework for sensor estimation, a critical step of many machine condition monitoring systems. The temporal behavior of normal sensor data is described by a stationary switching autoregressive (SSAR) model that possesses two advantages over traditional switching autoregressive (SAR) models. First, the SSAR model removes time dependency of signals during mode switching and fits sensor data better. Secondly, the SSAR model is stationary in that at each time, sensor data have the same distribution which represents the normal operating range of a system; this ensures that estimates are accurate and are not distracted by deviations. During monitoring the deviation covariance is estimated adaptively, which effectively handles variable levels of deviations. Tests on gas turbine data are presented.

Character segmentation algorithm for recognition of vehicle license plate

In the automated license plate recognition system, many reading errors are caused by inadequate character segmentation. In particular, character segmentation becomes difficult as the acquired vehicle images are seriously degraded. In this paper, we use computer vision techniques and propose a recognition-based segmentation method coupled with template matching and neural network. This will enhance the accuracy of the recognition system that aims to read automatically the German license plate. Algorithmic improvements for a projection-based segmentation are described here. In the preprocessing stage, plate tilt detection and position refinement methods are developed to prepare the data for later process. For separating touching characters, a discrimination function is presented based on a differential analysis of character contour distance. A conditionally recursive segmentation with the feedback of recognition is developed for effectively splitting touching characters and merging broken characters. We have implemented our algorithms in the intelligent camera system and obtained improvement for the recognition rate. Currently, the experiment conducted with greatly varying illumination conditions is shown at the recognition rate of an average of 92 %. Further improvement of the system is continuously undertaken for various data sets acquired under different environment conditions.