【正文】 he storage of a multiband raster dataset. In the future this XML file and the corresponding schema could be supplied directly by the provider of the remotely sensed data. In other words, the initial data could be an XML file including the data and the metadata of the scene along with the XML schema, which will be the reference for each scene instance of each satellite sensor. A sample of the XML file is shown in Fig. 9. ?xml version= encoding=39。 row=39。 b171/b1 b221/b2 b317/b3 b49/b4 b56/b5 b75/b7 /pixel pixel id=39。 column=39。 row=39。 b171/b1 b221/b2 b316/b3 b49/b4 b56/b5 b75/b7 /pixel ... ... /image Fig. 9. Sample of XML file. Metadata storage is an issue that must be considered depending on the application. Metadata about the remotely sensed image may include a variety of information ranging from pixel size and the number of rows and columns, up to the condition of the sensor during the capture of the scene and orbital data. It can be stored in the same file with the pixel values or in a separate file using the XML format. Both ways are acceptable since XSLT is able to manipulate multiple documents (. data and metadata file) as input. Following the same procedure, the ASCII file can be transformed directly to SVG image file. This structure of the SVG image file can take advantage of the nested structure supported by the SVG specification and nest all bands in a single SVG file. Since it is possible to show only three bands at a time, the choice of the bands to be shown along with their order (color/band correspondence) could be decided by the user with a set of predefined widgets written in ECMAscript. This is the equivalent of creating pseudocolored images with proprietary remote sensing software. . Classification The result of the process described in can be transformed into a classified XML file again with the use of XSLT. A number of XSLT templates are applied depending on the entities to be classified. More specifically, the objective is to classify vegetation, waste land and water regions using two normalized band ratios. The first band ratio discriminates water from land and the second is a vegetation index. These ratios take into consideration the values of two bands (1 and 5 the first and 3 and 4 the second) of each pixel and assign to it the new value that is calculated through the following formulas: Sea_indicator = (b1 b5)/(b1 + b5), (1) Vegetation/Waste land_indicator = (b4 b3)/(b4 + b3). (2) The second formula is widely known as Normalized Difference Vegetation Index—NDVI (Lillesand and Kiefer, 2021). The oute of this procedure is an XMLencoded file, which describes an array of pixels whose values result from formulas (1) and (2). Both formulas are inserted as templates in a XSLT stylesheet (Fig. 10), which transforms the XMLencoded image file into a classified one (Fig. 11). ?xml version= encoding=UTF8? xsl:stylesheet version= xmlns:xsl= xsl:output method=xml version= encoding=UTF8 indent=yes/ xsl:template match=/ image xsl:foreach select=/image/pixel pixel xsl:attribute name=id xsl:valueof select=id/ /xsl:attribute xsl:attribute name=row xsl:valueof select=row/ /xsl:attribute xsl:attribute name=col xsl:valueof select=column/ /xsl:attribute sea_ind xsl:valueof select=substring((b1 b5) div (b1 + b5),1,5)/ /sea_ind vw_ind xsl:valueof select=substring((b4 b3) div (b4 + b3),1,5)/ /vw_ind /pixel /xsl:foreach /image /xsl:template /xsl:stylesheet Fig. 10. Creation of band ratios with XSLT. ?xml version= encoding=UTF8? image pixel id=0 row=0 col=0 sea_ind/sea_ind vw_ind/vw_ind /pixel pixel id=1 row=1 col=0 sea_ind/sea_ind vw_ind/vw_ind /pixel pixel id=2 row=2 col=0 sea_ind/sea_ind vw_ind/vw_ind /pixel ... ... /image Fig. 11. Sample of classified image file. The most efficient way to visualize geographic information stored in XML is through SVG. The XSLT file includes a template that supports one of the abovementioned parisons. The parison results in a trueorfalse state and accordingly a black or white value is assigned to each pixel position resulting in a ‘‘binary’’ SVG file that has a rect element for each pixel filled with black or white color. The size of the rectangle can be set either to 1 1 pixels or to 25 25m (the actual size of the corresponding ground element) depending on the requirements of the application. This process results in a classified SVG image file, which differentiates sea (black—background) from land (white—foreground), as shown in . shows the oute of the classification resulting through ERMapper, which is identical with the SVGencoded file. In order to achieve the classification of the sea/land the same steps are followed. It is pointed out that these methods do not constitute a sophisticated implementation of remote sensing techniques. They only prove the concept of applying basic remote sensing and image processing principles utilizing open standards and XMLbased technologies. Fig. 12. (A)Classification through XML technology. (B)Classification through ERMapper. A subset of the classified XML image file is shown in Fig. 11. If sea_ind the pixel corresponds to the sea, otherwise it denotes land. In the same way if the vw_ind then the pixel belongs to vegetation. . Raster to vector conversion The conversion of the SVG image file into vector may sound awkward since the elements contained in a SVG file are already vectors. Well, in this case a vector element (the rect element) is used to describe the data of a raster dataset. In other words the rect element bridges the raster and the vector. Information encoded in the SVG file has to be decoded