A STOCHASTIC PROPOSAL TO EVALUATE CHANGES IN SAR IMAGES
Images SAR. Change Indicator Matrix. Stochastic distances, modified segmentation difficulty, contrast.
The detection of phenomena on the earth's surface is an important and necessary work, especially when it comes to large-scale phenomena such as oil slicks in the sea and deforestation in large forests. However, the monitoring of these phenomena through optical satellite images brings with them complications in terms of continuity in image generation. The Synthetic Aperture Radar, however, presents itself as ideal for monitoring the earth's surface reliably, continuously and globally, as it allows the generation of images with its own light source and with a higher resolution than conventional optical systems, in addition to enable its operation regardless of weather conditions. However, the difficulty in working with these images is centered on the interpretation of imaged data, requiring processing to extract information. This is due to the presence of noise of a stochastic nature that significantly degrades the image. Thus, the use of probability distributions is of fundamental importance in modeling these data. In this dissertation, we propose a measure of contrast for SAR images that presents significant gains in terms of information and classification of SAR images, which can be used for different applications.