Análise de remanescentes florestais utilizando classificação orientada à objeto através da lógica fuzzy no município de Ponte Alta do Norte – SC

Abstract

New technologies such the generation of images obtained by remote sensors has been essential in environmental studies. Thorough them, the more distant environments or hard to become more accessible. Using images from CBERS, SPOT and orthophoto studies in remnants of native forest in Ponte Alta do Norte - SC. The study was discussed in two chapters. Chapter One was aimed at analyzing the phytosociological parameters of the horizontal structure of the study area, in this case, a Dense Ombrophilous Montane Forest. These remnants were sampled using the sampling point away from the plant, represented by the quarter method. Were installed 134 sites in seven transects. We measured 536 individuals belonging to 59 species, 35 genera and 23 families. The area sampled was stratified according to the structure of the remnant forest, where the cluster analysis was performed by the index of Bray-Curtis dissimilarity. Formed four vegetation groups (four strata). The Stratum 1 showed a face in a more advanced stage of succession. However, the second stratum was defined as an area in the early stages of succession. The stratum 3 showed considerable density of two endangered species, Dicksonia sellowiana and Ocotea porosa. Strata 3 and 4 had intermediate stages of succession. The forest is in good condition because most of the strata is tightly structured. Because they occur two endangered species in a survey, the remnant forest inserted in Rio das Pedras farm can be considered a High Conservation Value Forest. Chapter II was aimed at making an object-oriented classification using fuzzy logic in digital images of three different forest areas to quantify the expression of the study area, as High Conservation Value Forest. As a methodology for the extraction of spectral information in the images were used the analysis techniques for object-oriented image classification. We made a comparison between different values to determine the segmentation, followed by identification of parameters that best discriminate the classes studied (strata). Sampling transects are carried out in Chapter I generated information for the development of membership functions, which were standardized and developed based on the parameters of the value of brightness and texture of objects sample. The membership functions developed for each class were crossed by operators of fuzzy logic, the algorithm formed then was generated image classification. To analyze the accuracy of classification was developed in field samples truths which were cross-classified and analyzed by Kappa. The results were compared to the three images, and the best result was for the image segmentation Orthophoto multi-resolution level, which was used to classify and quantify the remaining forest, totaling 133.02 ha in area which should be managed and maintained as a High Conservation Value Forest, according to characteristics raised in the study of phytosociological parameters of the horizontal structure (Chapter I).