Desarrollo de microorganismos Desarrollo de una librería en código libre para clasificar imágenes hiperespectrales por mapeo del ángulo espectral
Fecha
Autores
Autor corporativo
Título de la revista
ISSN de la revista
Título del volumen
Editor
Compartir
Director
Altmetric
Resumen
Discovering the interrelation between color, frequency and wavelength by William Herschel, the principles used to characterize incidence of light on objects [5], fundamental principles of remote sensing. Remote sensing has become an essential tool for the study, monitoring and understanding of phenomena not only of the earth's surface but also of the celestial bodies. Remote sensors have made great strides, from the very first panchromatic cameras used for aerial photogrammetry, multispectral scanner (transported by satellites), to the CASI sensor (with a spectral resolution of 288 bands [6]). This progress is reflected in your product "Image", which increased the amount of data stored, from multispectral to hyperspectral; thereby allowing for improved accuracy in the capacity for analysis and synthesis, however, this demands better capabilities both software and hardware to perform the processes [37]. Within the different applications Software specialized in processing data of remote or raster sensors are known Erdas, Envi, Ilwis, Grass, among others; the first two are commercial and the last one is free software. The most that allow these software are: the manipulation of raster information and vector, georeferencing and segmentation of images, spectral analyzes and image classification [11] [35]. Raster data or images are in different formats and types; the types currently used are multispectral, but since the In the past decade, the analysis of hyperspectral images has been very active. The hyperspectral image is the product of the technological development of spectral sensors [37]. The classification of the images is a task that is done with the purpose of quantitative data to qualitative [29], to achieve this there are different methods that have been implemented in this complex task, the most relevant are: maximum likelihood, Isodata, Fuzzy and, within the algorithms of intelligence artificial are retro-propagation and tree-making [25] [8]. The type of image determines the form of classification, for images the most prevalent methods are: Spectral Angle Mapping, Analysis of Spectral Mixture, Analysis of the whole Pixel, Montage characteristic Spectral Analysis, Sub-Pixel Analysis, among others. Grass software has few tools for image processing hyperspectral, especially that allows the classification by Mapping of the Spectral Angle (SAM, Spectral Anguler Mapper). Generating the need to develop a library that allows this classification in the software, migrating the SAM equation to a programming language supported by Grass to run on your kernel; which will allow users to access a free software tool to perform the classification SAM.