Machine vision in experimental poultry inspection

Machine vision in experimental poultry inspection

N DT Abstracts A neural network with an analog output is presented to determine the angle of inclination of a surface-breaking cracking from ultrasoni...

147KB Sizes 2 Downloads 92 Views

N DT Abstracts A neural network with an analog output is presented to determine the angle of inclination of a surface-breaking cracking from ultrasonic backscattering data. A neural network estimates the angle of inclination of the surface-breaking crack, assuming that the depth of the crack is 2.0ram, by utilizing the waveforms of backscattered signals from the crack. The plate with a surface-breaking crack is immersed in water and the crack is insonified from the opposite side of the plate. The angle of incidence with the normal to the insonified face of the plate is taken to he 18.9sup(o). The neural network is a feed-forward three layered network. The theoretical data obtained by the boundary element method are used for the training. The performance of the trained network is tested by synthetic and experimental data.

Koido, J. Covering thickness and diameter measurement of reinforcing bars by eddy current testing using neural network 60119

Review of Progress in Quantitative Nondestructive Evaluation, Snowmass Village, Colora6o (United States), 31 J u l . - 5 Aug. 1994. Vol. 14A, pp. 841-847. Edited by D.O. Thompson and D.E. Chimenti. Plenum Press (1995) ISBN 0-306-45062-3. Techniques for determining the covering thickness and diameter of reinforcing bars are needed in the evaluation of the existing strength of reinforced concrete structures. Neural networks were employed to estimate covering and diameter of reb,ur in this study. The phase waveform of the eddy signal generated by scanning the test coil along the surface of concrete was used for input data for the neural network.

59990 Glinkowski, M.T.; Wang, N.--C. A new fault location technique for underground distribution circuits using artificial neural network Proceedings of the American Power Conference, Spons. Illinois Institute of Technology, Vol. 57-1, pp. 716-719 (1995) ISSN 0097-2126 A novel technique for locating faults in underground cable circuits is presented. The technique utilizes a concept of artificial networks. During a system fault a set of measurements of voltages and currents at various locations in the distribution circuit is used as a "pattern" that is applied to a specially-trained neural network (NN). This pattern is the compared with the library of training patterns (cases) to recognize the location of the fault. The output of the neural network is graphically displayed as a simple 3-D chart that provides an operator with an instantaneous indication of the location of the fault. The last section of the paper illustrates sample results for two typical distribution circuits and discusses the sensitivity analysis of the new method.

60118 Xu, J.H.; Birx, D.L. Neural network based paLttern recognition for defect detection of load/lock slots Review of Progress in Quantitative Nondestructive Evaluation, Snowmass Village, Colorado (United States), 31 Jul. - 5 Aug. 1994. Vol. 14A, pp. 835-840. Edited by D.O. Thompson and D.E. Chimenti. Plenum Press (1!995) 1SBN 0-306-45062-3. The minimum size of the crack required to he detected is 0.005" x 0.010". The challenge here is the changing scan surface length due to the elliptic shape and the small size (the minor axis is about 1/8" long) of the inspection area. Some of the :scan waveforms are depicted. We worked in frequency domain. 60036 Yim, J.; Udpa, S.S.; Udpa, L.; Mina, M.; Lord, W. Neural network approaches to data fusion Review of Progress in Quantitative Nondestructive Evaluation, Snowmass Village, Colorado (United States), 31 Jul. - 5 Aug. 1994. Vol. 14A, pp. 819-826. Edited by D.O. Thompson and D.E. Chimenti. Plenum Press (1995) ISBN 0-306-45062-3. This paper presents some new neural network based approaches for combining information obtained using multiple inspection methodologies in a synergistic manner to obtain more comprehensive information about the condition of the test specimen. Two specific application examples, one involving an attempt to fuse eddy current and ultrasonic images, and the other to fuse multifrequency ,eddy current images are described. Networks that were evaluated for iraplementing the fusion algorithm include multi-layer perceptrons (MLP) as well as radial basis function (RBF) networks.

59982 Schmoldt, D.L. Neural network classifiers to grade parts based on surface defects with spatial dependencies Review of Progress in Quantitative Nondestructive Evaluation, Snowmass Village, Colorado (United States), 31 Jul. - 5 Aug. 1994. Vol. 14A, pp. 795-802. Edited by D.O. Thompson and D.E. Chimenti. Plenum Press (1995) ISBN 0-306-45062-3. This report details initial efforts to develop a real-time lumber grader. Classifier development is introduced conceptually, spatial features are described, and the use of training and testing data sets is discussed. A utility function for comparing different classifiers is derived and demonstrated. Comparison tests using three different ANN topologies indicate where improvements can he made in classifier design and in training.

60034 Brown, L.M.; DeNale, R. Knowledge-based NDE system Review of Progress in Quantitative Nondestructive Evaluation, Snowmass Village, Colorado (United States), 31 Jul. - 5 Aug. 1994. Vol. 14A, pp. 787-794. Edited by D.O. Thompson and D.E. Chimenti. Plenum Press (1995) ISBN 0-306-45062-3. The objective of this work was to develop an automated ultrasonic system capable of locating, sizing, and classifying discontinuities located in hull welds. In addition, this system was developed to perform a computer implementation of the UT acceptance criteria requirements defined in NAVSEA 0900-LP-006-3010.

Pellerin, C. Machine vision in experimental poultry inspection 59918

Sensor Review, Vol. 15, No. 4, pp. 23-24 (1995) This article presents data from research based on colour vision, neural networks and precursor image processes as neural net input. Results show that neural net inputs of histogram values based on colour intensity, hue and saturation are good for detecting bruises and skin conditions; and that multichannel texture-analysis outputs are good for detecting skin tear candidate regions. These results also should he useful to automators in the fruit, textile and lumber industries.

60033 Zgonc, K.; Achen:bac'h, J.D.; Lee, Y.C. Crack sizing using a neural network classifier trained with data obtained from finite element models Review of Progress in Quantitative Nondestructive Evaluation, Snowmass Village, Colorado (United States), 31 Jul. - 5 Aug. 1994. Vol. 14A, pp. 779-786. Edited by D.O. Thompson and D.E. Chimenti. Plenum Press (1995) ISBN 0-306-45062-3. We discuss a neural network designed for use in ultrasonic signal classification. The network ~-an give classification results in a short time which makes possible real time ultrasonic inspection. The latest improvement is the use of numerically obtained ultrasonic data to train the neural network classifier (NNC).

59821 Okure, M.A.E. Neural networks for NDE signal classification Dissertation Abstracts International, Vol. 56, No. 3, p. 1668-B (Sep. 1995) (DA9521783) This theses addresses some of the fundamental problems encountered in the development of neural networks for NDE applications.

Takadoya, M.; Yabe, Y.; Kitahara, M.; Achenbach, J.D.; Guo, Q.C.; Peterson, M.L. An artificial intelligence technique to characterize surfacebreaking cracks 60032

Review of Progress in Quantitative Nondestructive Evaluation, Snowmass Village, Colorado (United States), 31 Jul. - 5 Aug. 1994. Vol. 14A, pp. 771-778. Edited by D.O. Thompson and D.E. Chimenti. Plenum Press (1995) ISBN 0-306-45062-3.

331

59818 Ceravolo, R.; De Stefano, A.; Sabia, D. Hierarchical use of neural techniques in structural damage recognition Smart Materials and Structures, Vol. 4, No. 4, pp. 270-280 (Dec. 1995) This study explores the possibility of using neural techniques to detect the presence of structural faults from dynamic response data in simple structures of practical interest for structural engineering. Some techniques of structural response signal processing and analysis devised for the purpose of improving the diagnostic capabilities of connectivistic methods are proposed through numerical experiments. To this end, cross-correlations between dynamic response signals are used as inputs for hierarchically organized networks which are able to assess and locate structural damage in numerical models.