| Popis: |
In this paper a non-linear correlation methodology to recognize objects is used. This new system is invariant to position, rotation and scale. This digital system has a low computational cost to achieve a significant reduction of processed information by using vectorial signatures. The invariant vectorial signatures are obtained from the information from both the target image as well as problem image. In this way, each image has its rotational and scale vectorial signature obtained through several mathematical transformations such as scale and Fourier transform. So, this method uses the great capacities from the non-linear filters to discriminate between similar objects. Vectorial signatures are compared using non-linear correlation. The result of this comparison is shown in a bi-dimensional plane where the x axis is the result of the rotation correlation and the y axis is the result of the scale correlation. In addition, spectral index and vectorial signature index are obtained through several mathematical transformations in order to recognize the objects in a more simple way. 21 different fossil diatoms images were used. The results obtained are analyzed and discussed. |
