Abstract
This paper describes a practical active vision system called Sheila that equips two differentiated visual fields. Sheila detects insulator images with a peripheral vision and shifts its fixation point sequentially to each object for examining them precisely. Insulator images in windows expose characteristic peaks in the local frequency domain. Such patterns are detected by two sorts of cell-detectors, the periodicity cells and the directionality cells. The output signals of the cells are transmitted to a neural network that stimulates where insulators exist in the image. Experiment results for a transformer substation showed a practical potential of Sheila.
