Automatic Stereoscopic Recogniton of Observed Scences

摘要

The possibility of stereoscopic recognition of observed scenes using their intensity and geometrical patterns on the base of the information contained in a pair of stereo images acquired by an optical electronic iconic system is investigated. The algorithm of automated processing is based on the method of adaptive hypothesis recognition. The functional being minimized has two terms: a quadratic form of the differences between the measured intensity of various points within the scene and their reference intensity and a quadratic form of the differences between the intensities of the conjugated points of the scene in the first and the second images of the stereoscopic pair. As an example, the problem of stereoscopic detection of the plane that is elevated over the surrounding background is solved. The relation between the probability of error of detection and the signal noise ratio, the height and size of the object, the ratio of the variation of intensity for the object to the error of the receiver and the resolution of the optical system is obtained.