A bit of remote sensing for rock paintings

0
993

One of the latest tools I have encountered belongs to the field of remote sensing, understanding this discipline as the means of acquiring information about an object without establishing contact with it.

Currently, these technologies are almost exclusively used in projects related to environmental, meteorological aspects, etc., thanks to the proliferation of satellites equipped with multispectral vision equipment among other technologies.

However, many characteristics of this discipline are relatively interesting to serve as a complement to digitization and diagnosis projects on works of art, archaeological artifacts, or rock paintings.

The potential of photography in remote sensing lies in the multispectral dimension of the images taken with these devices; unfortunately, these types of cameras are especially inaccessible due to their costs, so I will make a brief review of these techniques using traditional RGB photography.

The software to be reviewed in this article will be MultiSpec, which is a free tool sponsored by Purdue University. MultiSpec, like all tools of this type, supports a multitude of multispectral or hyperspectral images, the main characteristic of this type of images being that they have several channels, where captures related to a particular wavelength are successively stored. In our digital cameras, depending on the type of development, we usually have three channels R, G, and B, each with different densities.

remote sensing rock artMultiSpec will allow us to work with color images, that is, with the three channels or choose one of them. When working with TIFF images resulting from a RAW development, the colors are the result of the demosaicing algorithms necessary to convert our Bayer matrix into a color image. If we want to work with a particular channel of our images, it is more appropriate to develop them discarding the demosaicing; this can be done with tools like libraw and its utility 4channels, so that we obtain four images: R, G, G2, and B, where each corresponds to the wavelength range with the Red, Green, and Blue photoreceptors of our Bayer sensor.

Among the functions available in MultiSpec, perhaps the most characteristic are classification tasks, understood as the tasks of organizing an image using a false color map, by areas that present a similar density. This is achieved through a training phase, where we describe which densities we want to map; in this way, the application will extrapolate the density patterns we have proposed to the rest of the image. These processes are performed on the histogram; in fact, it is a very typical tool in this type of application.

The workflow with MultiSpec to carry out a classification project could be as follows:

pantalla_inicio1.- After activating the menu to load a new image, we will be asked what type of image we are opening; normally, we will choose Channel Color or Grayscale.

2.- Once the image is opened, we must go to the Processor menu and activate the Statistics option (command+t); now we are ready to create the training samples.

teledeteccion arte rupestre3.- We proceed to define representative areas in our image; the default cursor tool is simple rectangles, however, from the project panel we can activate the option to create polygons, so that we can create areas with more complex shapes.

nombre_muestras4.- After creating a sample, the Add to list option is activated, so we can load and describe it correctly. Samples are organized into Classes and Fields, so that a class can be organized into several fields.

5.- Once the necessary number of samples has been created, we return to the Processor panel and activate the Classify option. In this panel, we can make decisions about the type of procedure to use or the channels we will use. However, the first option we should familiarize ourselves with is the one under the term “Write classification results to:” and then activate “Disk file”; in this way, we will create a .GIS file, which is our image with a false color mapping of the samples we described in the previous process.

classify_menu

6.- Once the .GIS image is created, we can proceed to load it. In the loading options for this type of image, we can choose which classes we want to visualize, etc.

If instead of performing this training process, we directly open an image with the “Channel Thematic” option, a mapping of the image densities is done directly. This is a fairly quick option to obtain a classification of the elements of an image based on the similarity of densities.

Beyond the simplicity of this process, we can complicate it by making decisions about the channels involved, types of algorithms, etc.

digitalizacion, teledeteccion arqueologiaAs we see, it is a fairly quick process that can help us perform classifications or mappings on our images based on the densities of their channels or their colors. Classification techniques with remote sensing tools can be useful for creating digital drawings conditioned by certain aspects of photographs, tasks very common in the fields of archaeology, rock art, or cultural heritage conservation.