Estimación de la altura de columnas explosivas en el volcán Tungurahua basado en procesamiento de imágenes visuales y térmicas.

Abstract

The objective of the present research was to estimate the height of the eruptive column of Tungurahua volcano, based on the visual and thermal image processing through the application of techniques and algorithms of artificial vision. The height of the eruptive column is a parameter in the calculation of the Volcanic Explosivity Index (VEI). It is a relative measure to determine the magnitude of the volcanic eruptions. The Tungurahua Volcano Observatory (TVO) facilitated the thermal images for its processing. Photointerpretation techniques

were applied and 5 algorithms of test in MATLAB (matrix laboratory) were implemented based on mathematical morphology, decomposition RGB color channel and HSV filter. The results of each algorithm were evaluated by observation with 12 test images. As a percentage of the attenuation of the atmospheric elements was weighted between the raw image and the processed image. It was selected the algorithm that obtained the most significant measure in the attenuation of atmospheric elements, 14.58%. This algorithm used an HSV filter for red and oranges tones in the range of 352 to16 degrees; the decomposition of RGB channels to attenuate any element other than the eruptive column; the morphological process of geodetic expansion to obtain only the edges of the column. Finally, a relation of 9.0909 meters per pixel was estimated in the image for the digital measurement of the column from the edge of the crater. It was concluded that through the use of artificial vision techniques, the processed thermal image shows a significant difference in the attenuation of atmospheric elements and the eruptive column. The height is estimated by a calibrated digital rule that allows horizontal, vertical and oblique measurements on the eruptive column in the MATLAB programming, facilitating VEI calculation for Tungurahua volcano monitoring and Risk Management in Ecuador.