| Abstrakt: |
The problem of classifying mobile objects present on the stage according to the shape of their projections is considered. During the construction of the projection of the object on the image of the scene, it can lose some of its pixels, and add some pixels from its environment. As a result, the shape of the projection of an object built from the image may differ from the ideal projection built from its 3d models. The paper proposes a quantitative measure of projection distortion and determines its influence on the probability of correct classification of the distorted projection shape. It is obvious that the quality of the constructed projection, measured by the value of its distortion, depends on the combination of various properties of the image. For their designation, the word complexity is used below. Naturally, the increase in complexity increases the distortion of the projection and reduces the probability of correctly classifying its shape. The paper proposes a quantitative measure of image complexity and establishes the dependence on it of the probability of correct object classification. The results obtained in the work allow, on the one hand, to assess the probability of correct object classification depending on the complexity of the image without classification. On the other hand, they allow us to formulate requirements for the complexity of the image for a given probability of correct classification. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink