Learning Descriptions of 2D Blob-Like Shapes for Object Recognition in X-Ray Images: An Initial Study
Marcus A. Maloof, Ryszard S. Michalski · 1994
This paper describes a method for applying AQ15c to learning shape descriptions of 2D bloblike objects in x-ray images. The methodology and initial experimental results are discussed, along with comparisons to k-nearest neighbor and to feed-forward neural networks. The AQ15c learning method is shown to have distinct advantages over the aforementioned techniques in terms of higher or comprable classification accuracy, learning and recognition time, and understandability of learned concepts. This approach is well-suited for recognizing objects that can be isolated in the image using histogram and thresholding techniques and that have little internal structure. Key words: machine learning, machine vision, shape recognition, concept learning. Acknowledgments The authors would like to thank the many people whose useful discussions contributed to this work: Jerzy Bala, Eric Bloedorn, Ibrahim Imam, and Ali Hadjarian. The authors wish to also thank those who read preliminary drafts of this p...