Search

This paper proposes the use of deep learning classification for acoustic monitoring of an industrial process. Specifically, the application is to process sound recordings to detect when additional air leaks through gaps between grate bars lining the bottom of the sinter strand pallets, caused by thermal cycling, aging and deterioration. Detecting holes is not possible visually as the hole is usually small and covered with a granular bed of sinter/blend material. Acoustic signals from normal operation and periods of air leakage are fed into the basic supervised classification methods (SVM and J48) and the deep learning networks, to learn and distinguish the differences. Results suggest that the applied deep learning approach can effectively detect the acoustic emissions from holes time segments with a minimum 79% of accuracy.

We employ Eigenfaces to discriminate between handwritten and machine-printed text at the connected component (CC) level. Normalized images of machine print CCs are treated as points in a high-dimensional space. PCA yields a reduced-dimensional character space. Representative machine print CCs are projected into character space and a local distance threshold for each representative is automatically determined. CCs are classified as machine print if they are within the local distance threshold of their closest machine print representative. Otherwise, they are classified as handwriting. Recursive character segmentation using min graph cut is used to address the problem of touching characters. Validation over a large NIST handwriting and machine print database demonstrates precision of 93.98% and 89.1% for machine print and handwriting respectively.

The problem of locating a reference image within a larger image using a correlation technique is discussed. Although the particular application discussed is that of locating a reference image obtained from one video sensor or a photograph, within the larger field of view obtained from a second video sensor in real time (i.e., at the TV field rate), the results are general and useful for a number of applications. The tradeoffs necessary to obtain real time correlat are discussed and their effect on correlation accuracy is given.

A brief review of the principles of stereoscopy is given followed by a discussion of candidate methods for providing stereoscopic viewing of display data. An analog computer implementation of a stereoscopic graphics display system is described. The analog implementation provides real-time display capability but has no data storage capability. Implementation with a small digital computer provides data storage but is significantly slower for displaying dynamic changes in the data. Other points of comparison between the two systems are given with the conclusion that a hybrid system would combine good features of both. Experimental evaluation relative to eye fatigue is given for three specific viewing techniques. Example stereograms are given for both the digital and analog systems. Of particular interest was application of the display system for viewing biomedical data, as illustrated by an example display of a vectorcardiogram.

Electroencephalographic (EEG) data consisting of visual evoked responses monitored via an array of electrodes from humans and penicillin-induced focal epileptic discharge data recorded from rat neocortex are analyzed. This procedure, previously applied to seismic array data, offers a method by which high-resolution vector velocity and, thus, direction and speed of propagating wavefronts can be estimated.

A practical system for computing attitude angles from the image of four rectangular, coplanar points has a number of potential error sources that degrade the accuracy of the attitude computation. An error analysis is presented for the Robot Locating System (RLS), which is a multisensor laboratory testbed that seeks to compute attitude angles from an image of four points. This paper discusses potential error sources and their relative effects for the RLS system. The error sources are discussed from a physical point of view and expected error characteristics are described. These error sources are then evaluated using a software simulation of the RLS system.

Two iterative algorithms are presented for phase retrieval from intensity measurements for two-dimensional complex-valued sequences. The first algorithm constructs a linear combination of present and past iteration outputs using a relaxation parameter. The second algorithm utilizes a Kalman-type filter to combine estimates of object domain magnitude. A new error metric is defined that uses residual monitoring to show the relationship of the error to the reconstructed image quality. These algorithms are first tested against real-valued images to investigate their convergence chacacteristics. Both algorithms are then applied to infrared images which are characterized by low-contrast and noise-corruption. The question of uniqueness is also addressed for two-dimensional complex-valued sequences.

The object-oriented paradigm offers a new approach to the development of software for image processing applications. The fundamental concept of the object-oriented paradigm is that problem solutions are implemented by sending messages to objects. This requires that the objects in a solution be defined along with messages to which each object will respond. This is in contrast to structured programming wherein one defines data structures and sends those data structures to a procedure as parameters.