Geodeziya, kartografiya, geografiya

What about preprocessing?
Preprocessing can be divided into two categories; radiometric and geometric processing. Radiometric processing is used to reduce the effects that prevent us from measuring surface reflectance using remotely sensed imagery. This can involve reducing the effects the atmosphere has on modifying solar energy (sunlight) before and after it hits the Earth's surface and the effects due to terrain and sensor configurations. Minimizing this "noise" reduces variation in the value of a pixel due to factors independent of surface cover. When conducting automated classification, reducing these effects tends to improve classification accuracy since you are essentially reducing the noise in the image. With visual interpretation, however, this is not always the case. For example, in mountain environments there is often one side of a mountain that is brightly lit while the other side is in shadow. If not corrected, this effect will introduce problems for automated classification since the same land cover will look very different.
However, with visual interpretation this illumination effect can actually help since it tends to accentuate 3-dimensional features. Moreover, a trained analyst familiar with the study area is able to accurately classify and identify the vegetation even though the shadows and topography make the vegetation features appear like two distinct classes.
In the past, using radiometric correction methods required a significant level of expertise and data that was not widely available. In recent times, however, radiometric correction algorithms have become more accessible to the non-specialist and as this trend progresses it is likely that these tools will be able to improve significantly the ability of non-specialists to benefit from these corrections.
Geometric correction typically involves warping an image to match a specific map projection. Whenever an image is warped a resampling of the image pixels takes place which degrades the original data to some degree. This has more of an influence when applying automated methods since the individual pixel values are modified during resampling. This effect may or may not be significant, depending on the application, but in general it is best to perform the geometric corrections after the final classified map is produced. This is not an issue when doing visual interpretation because the changes in pixel value due to resampling are not as visible to the human eye.