An application of neural networks to natural scene segmentation
This paper introduces a method for low level image segmentation. Pixels of the image are classified corresponding to their chromatic features.
The classifier is implemented by means of a neural network trained using a manual clasification of the coloured images. The back-propagation algorithm is applied to one image which works as training set, and the results applied to the rest of them.
Because of the large amount of data contained in one image, a random selection of the points is made before training.
Finally, the paper shows the results of applying the method to fruits detection in natural scenes taken from the open field.
Unable to display preview. Download preview PDF.
- Haralick, R. M. and Shapiro, L.G. "Image segmentation techniques". Computer Vision, Graphics and Image Processing, 29 (1985).Google Scholar
- Ballard, D.N. and Brown, C.M. "Computer Vision". Prentice Hall, 1982Google Scholar
- Wyszecky, G. and Stiles, V.S. "Color Science" John Wiley and Sons Inc, 1978Google Scholar
- Pratt, W.K. "Digital Image Processing" John Wiley and Sons, Inc. 1978Google Scholar
- Albert, J. Ferri, F., Domingo, J. and Vicens, M. "An approach to Natural Scene Segmentation by means of Genetic Algoritms with Fuzzy Data".Google Scholar
- Lorentz, G.G. "The 13th. Problem of Hilbert" in F.E. Browder (Ed). Mathematical developments Arising from Hilbert Problems, American Mathematical Society, Providence, R.I. 1976.Google Scholar
- Kohonen, T. "Self-Organization and Associative Memory". Springer-Verlag, Berlin, 1984 and 88Google Scholar