A Pre-Attentive Vision Model for Data Prospecting Ivan A. Galkin, Grigori M. Khmyrov, Alexander V. Kozlov, Bodo W. Reinisch, James C. Tilton, Shing F. Fung An increasing number of technical and academic applications rely on the intelligent systems for imagery data prospecting to cope with the information avalanche. An imagery data prospector pre-classifies input images in categories so as to reduce the unrealistic demand of the manual labor needed to analyze all of them. We studied applicability of the bottom-up pre-attentive vision models to the task of data prospecting. Generally, these models do not assume prior knowledge of the domain discipline and specific characteristics of signatures and merely implement the Gestalt perception principles to find features that stand out against the image background. We developed and tested such model against the RPI plasmagram dataset currently comprising 1.1 million images. The model implements a recurrent neural network that optimizes alignment of rotors placed on top of detected edge elements to identify whether the image contains contours of RPI signal reflections. The paper discusses model design, presents performance results, and outlines future efforts. _______________ Proceedings of the Conference on Cybernetics, Information Technologies, Systems, and Applications (CITSA), 2005