Estudio de la estructura anatómica que conforma la madera mediante análisis multiresolución y morfología

Abstract

The microanatomy structure of wood is composed by 4 types of cells: vessels or pores, the parenchyma, the spokes and the fiber. Its form, deposition and texture can tell us what the species to wich is it belongs. Its characteristics are studied through a sample taken from the outer part of the wood. The sample represents three planes, wich provide different information about the cells: transverse, tangencial and radial. These planes can be represented in photographs taken from a microscope to be treated as images and be processed. Some works have segmented elements to be able to identify wood with image processing. This work makes a study of a segmentation model, based on multiresolution analysis and morphological operations, using the transversal plane. Starting with the adaptation of the images and using gaussian filter, the decomposition of the image is done using wavelet transform to find the details that make the features and segment the cells. With morphological operations, noise and unwanted elements are eliminated, and finally the images are reconstructed using the wavelet inverse transform. Two sets of results were obtained for the evaluation. The first one included all the images, even those with defects in the sample or errors in the capture. The second one omitted those images which, according to our criteria, were not suitable for study. In comparison to the first set, the second had a more successful result, which mostly was a value greater than 80 % of the F1 Score, a measure that was used for evaluating the model.