GULLY EROSION DELIMITATION THROUGH GEOBIA (Geographic Object Based Image Analysis) AND DATA MINING

It is possible to see a certain amount of attention on the part of public agencies, with regard to mapping, monitoring areas, for a reliable environmental control. One justification for this statement is the creation of environmental programs by the federal government, such as: Rural Environmental Registry (CAR), Plan for the Recovery of Degraded Areas (PRAD). Both programs cover the mapping of areas in their structures. This reality makes it of the utmost importance to discuss better ways to obtain cartographic information about degraded areas. Soil degradations, of gullet type, present several damages to nature, as they have an irreversible state, being possible only their partial recovery. The monitoring of these areas, as well as information about them, is of paramount importance to ensure control and define conservation methods. Under this problem, this work aims to evaluate the efficiency of a classification procedure, oriented, semi-automatic (GEOBIA) in cartographic products produced by Remote Piloted Aircraft (ARP) for the delimitation of gullies. The use of cartographic products from ARP (Digital Elevation Model and Digital Orthography) is justified due to the low cost of the tool, as well as the planialtimetric potential. The procedures were carried out in two study areas, located in the city of Itajuba, MG. In these areas, some control and checking points were defined for the classification of cartographic products in relation to the Digital Cartographic Accuracy Standard (PEC-PCD). The products presented class A, for the scale 1 /2,000. Some segmentation parameters were determined to form reliable segments for each specific study area, then the most relevant attributes were determined for the classification and construction of the decision tree for each area. In making the decision tree, the C4.5 algorithm was used. The results were satisfactory at levels of precision (Kappa index between 0.92 and 0.89 and Global Accuracy of 93.99% and 91.18%, for areas 01 and 02 respectively), allowing the techniques used in cartographic products from ARP to be a tool for delimiting degraded areas of the gully type.