Flexible, reconfigurable vision systems can provide the richest sensing modality for sophisticated multiple robot platforms. We propose a cooperative and adaptive approach to vision applied to the problem of finding and protecting humans by a robot team in such emergent circumstances, for example, during a fire in an office building. A panoramic camera system plays an important role in this approach. This report presents the recent progress on the development of vision algorithms based on a panoramic annular lens (PAL) camera system. First, we give a brief survey of existing panoramic camera systems, with an emphasis on geometrical properties of capturing panoramic images from a single-viewpoint. Second, a mathematical model of the panoramic annular lens (PAL) camera that we use is built, and then several issues about camera calibration and image un- warping are addressed. An empirical method is given to roughly ''calibrate" the PAL camera system. Next, a panoramic virtual stereo (PVS) vision approach is proposed, and the problems of self-calibration and real-time 3D estimation are discussed. Finally, we present an experimental system that can detect and track multiple moving objects in real-time using a PAL camera system. The panoramic vision algorithm for moving object tracking also serves the initial step of the PVS approach.

Article Free Access Share on Indexing handwriting using word matching Authors: R. Manmatha Center for Intelligent Information Retrieval, Computer Science Department, University of Massachusetts, Amherst, MA Center for Intelligent Information Retrieval, Computer Science Department, University of Massachusetts, Amherst, MAView Profile , Chengfeng Han Center for Intelligent Information Retrieval, Computer Science Department, University of Massachusetts, Amherst, MA Center for Intelligent Information Retrieval, Computer Science Department, University of Massachusetts, Amherst, MAView Profile , E. M. Riseman Center for Intelligent Information Retrieval, Computer Science Department, University of Massachusetts, Amherst, MA Center for Intelligent Information Retrieval, Computer Science Department, University of Massachusetts, Amherst, MAView Profile , W. B. Croft Center for Intelligent Information Retrieval, Computer Science Department, University of Massachusetts, Amherst, MA Center for Intelligent Information Retrieval, Computer Science Department, University of Massachusetts, Amherst, MAView Profile Authors Info & Claims DL '96: Proceedings of the first ACM international conference on Digital librariesApril 1996 Pages 151–159https://doi.org/10.1145/226931.226960Published:01 April 1996Publication History 45citation594DownloadsMetricsTotal Citations45Total Downloads594Last 12 Months18Last 6 weeks5 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF

AbstractWe demonstrate the use of computer vision techniques and optical microscopy to follow the kinetics and microstructure during spinodal decomposition of a polymer blend. Among other features, the mean of the population of the local maxima of the gradients in each image is computed; this global feature is shown to co‐develop with the phase separation of the blend. An algorithm is presented which employs the gradient magnitude technique to analyze optical images of spinodally decomposing polymer blends. This algorithm has been used to extract the Cahn‐Hilliard spinodal growth rates for a binary blend of polystyrene with poly(vinyl methyl ether). We show that the spinodal temperature can be found from the temperature dependence of this growth rate. We also show how additional shape features such as compactness might be used to study, the same binary blend.

Relatively low character error rates can often lead to prohibitive levels of word error rates. This paper examines several techniques for integrating an independent contextual postprocessor (CPP) into a full classification system. Using positional binary n-grams the CPP can correct many errors directly. In those cases where the correction process leads to ambiguity, the CPP can direct additional processing. Experimental results demonstrate that almost all of the derived improvement results from CPP-directed reclassification. This only requires that the CPP have the classifier likelihood fed forward to it. Therefore, a standardized CPP can be built independently of the rest of the classification system. An initial 45% word error rate is reduced to about a 2% word error rate and a 1% reject rate. Presence of a dictionary allows these figures to be reduced even further.

This paper describes a special-purpose character recognition system which uses contextual information for the recognition of words from any given dictionary of words. Previous techniques that utilized context involved letter transition probabilities of digrams and trigrams. This research introduces the concept of binary digrams which overcomes some of the problems of past approaches. They can be used to extract offectively the "syntax" of the dictionary while requiring very modest amounts of storage. A computationally feasible procedure is described which allows the accuracy requirements of the character recognizer to be relaxed if it is followed by a contextual postprocessor. The modified recognition system is allowed to output several alternatives for each character, while the postprocessor selects the proper string of characters by having access to both the dictionary and the dictionary syntax. A theoretical estimate of the recognition rate is derived, and experimental results demonstrate the ability of the system to achieve low error and rejection rates.

In dynamic situations where the sensor is undergoing primarily translational motion with a relatively small rotational components, it might seem likely that approximate translational motion algorithms can be effective. It is shown quantitatively, however, that even small rotations can significantly affect the computation of the focus of expansion (FOE). This is shown theoretically for the case in which the environment is a frontal plane, and also experimentally. Two algorithms are presented. One is an existing general motion algorithm. The second is a pure translational algorithm based on the weighted Hough transform. The depth results obtained using the second algorithm are reasonable, although they are not as good as those using general motion.<>