Computer Vision through Learning
GEORGE MASON UNIV FAIRFAX VA
Pagination or Media Count:
The underlying motivation for this research is that vision systems need learning capabilities for handling problems for which algorithmic solutions are unknown or difficult to obtain. In this context, we have conducted research on a wide range of vision problems that can benefit from machine learning. We have developed a general methodology for this purpose, called MIST, that supports multilevel image sampling, transformation, learning and interpretation. MIST is based on the application of symbolic or multistrategy learning methods the latter one integrates symbolic and neural net learning for creating visual concept descriptions. MIST was applied and demonstrated to be useful for such problems as conceptual interpretation of natural scences, non-structural and structural texture description and identification, detection of blasting caps in X-ray images of airport luggage, and target detection in SAR images. We have also obtained important results in the areas of action recognition in video image sequences, visual memories, estimation of environment properties from sampling, and bisight head control. The obtained results have demonstrated a significant promise and usefulness of the efforts to apply modern machine learning methods to problems of computer vision.