Различение статистических свойств изображений медоносными пчелами

Abstract

Black-and-white figures, 100X100 elements in size, were synthethized by the digital computer. The probability of equal brightness of adjacent elements was set for calcula-tion. This probability determines some other statistics of the texture. Figures with the same parameters had the different situation of patches, formed by fused elements. The bees were trained to differentiate the figures (10X10 cm). In 42 out of 49 tests the bees made out the figures ‘of different statistical classes. Hense the statistical characters of patterns were the real clue for recognizing even where the actual form of the figure was occasional. When the difference between the figures diminished, the recognition became less reliable. The complex figures, combined from parts of different figures, distinguished from the homogeneous ones with the same average parameter, in 12 out of 12 tests. In 7 tests out of 10, the pairs of complex figures with the same statistics were distinguished. The bees can use a small plot (1-5 cm) with definite statistical properties on the back-ground with other properties as a pattern for recognition. The ability to distinguish such plots to the greater extent corresponds to an organized analysis of pattern in retina projections on optic ganglia, than to summarizing of signals from ommatidia independent in space.