Summary of scientific and technical accomplishments

* Developed neural network multispectral image classification ability and compared it to the standard maximum-likelihood statistical approach. It was found that the neural network approach is a good alternative to maximum-likelihood and may be more robust when faced with inferior training data, or multi-date or degraded imagery (such as with lossy compression - see the DIAL WWW page http://www-dial.ece.arizona.edu/CompClass/comp_class.html). The ability of the net to produce a useful classification with less training data is important in the context of a database search of single patterns. The idea is to be able to pick out a pattern quickly, train the search network, and then search through a database. The robustness of the classifier to the particular distribution of the search class is important for obtaining meaningful results.

* Developed parallel neural network training algorithm in C* on CM-2 and CM-5 massively parallel computers at NASA/Ames Research Center.

* PVM versions of neural network and maximum-likelihood classification algorithms developed for use on workstation clusters. Near optimal speedup, proportional to the number of (equal) workstations in the cluster, was achieved.

* Developed prototype system for identifying, training, and searching for a single multispectral pattern using the neural network approach. The major challenge here was that since the neural network training algorithm is competetive, suitable competing data for the single pattern of interest must be found or generated. It was found that a combination of random "background" data from the image and randomly generated patterns produced the best competing data set and allowed for the best disrcimination of single patterns. This is demonstrated on the DIAL WWW page at: http://www-dial.ece.arizona.edu/ImageSearch/image_search.html

Publications

Refereed journal articles:

* Paola, J. D., and Schowengerdt, R. A. (1995), "A Detailed Comparison of Backpropagation Neural Network and Maximum-Likelihood Classifiers for Urban Land Use Classification", IEEE Transactions on Geoscience and Remote Sensing, vol. 33, no. 4, pp. 981-996, July 1995.

* Paola, J. D., and Schowengerdt, R. A. (1995), "A Review and Analysis of Backpropagation Neural Networks for Classification of Remotely Sensed Multispectral Imagery", International Journal of Remote Sensing, vol. 16, no. 16, pp. 3033-3058, November 10, 1995.

Conference Proceedings

* Paola, J. D., and Schowengerdt, R. A. (1995), "The Effect of Lossy Image Compression on Image Classification", Proceedings, 15th Annual International Geoscience and Remote Sensing Symposium, Florence, Italy, vol. 1, pp. 118-120, July 10-14, 1995.

World Wide Web version with color imagery available at: http://www-dial.ece.arizona.edu/CompClass/comp_class.html

* Paola, J. D., and Schowengerdt, R. A. (1995), "Searching for Patterns in Remote Sensing Image Databases Using Neural Networks", Proceedings, 15th Annual International Geoscience and Remote Sensing Symposium, Florence, Italy, vol. 1, pp. 443-445, July 10-14, 1995.

World Wide Web version with color imagery available at: http://www-dial.ece.arizona.edu/ImageSearch/image_search.html

* Schowengerdt, R. A., and Paola, J. D. (1994), "Parallel Computing and Data Compression for Pattern Matching in Remote Sensing Image Databases", Proceedings, Conference on Recent Advances in Remote Sensing, The European Symposium on Satellite Remote Sensing, Proc. SPIE 2318, pp. 215-225, Rome, Italy, September 27-29, 1994.

* Paola, J. D., and Schowengerdt, R. A. (1994), "Comparisons of Neural Networks to Standard Techniques for Image Classification and Correlation", Proceedings, 14th Annual International Geoscience and Remote Sensing Symposium, Pasadena, Ca., vol. 3, pp. 1404-1406, August 8-12, 1994.

Technical Reports:

* Paola, J. D., and Schowengerdt, R. A. (1995), "The Effect of Neural Network Structure on a Multispectral Land-Use Classification", Research Institute for Advanced Computer Science (RIACS) Tech. Report TR 95.25, December 1995.

* Paola, J. D., and Schowengerdt, R. A. (1995), "Searching for Patterns in Remote Sensing Image Databases Using Neural Networks", Research Institute for Advanced Computer Science (RIACS) Tech. Report TR 95.17, August 16, 1995.

* Paola, J. D., and Schowengerdt, R. A. (1995), "The Effect of Lossy Image Compression on Image Classification", Research Institute for Advanced Computer Science (RIACS) Tech. Report TR 95.18, August 16, 1995.

* Paola, J. D., and Schowengerdt, R. A. (1993), "A Review and Analysis of Backpropagation Neural Networks for Classification of Remotely Sensed Multispectral Imagery", Research Institute for Advanced Computer Science (RIACS) Tech. Report TR 93.05, June, 1993.

Graduate students supported, advanced degrees awarded and thesis titles:

Justin D. Paola, Master of Science awarded in May 1994. Thesis title: "Neural Network Classification of Multispectral Imagery"

A non-html text version of this page can be found here