We propose a calculation procedure for obtaining multispectral images in remote sensing of vegetation objects in the spectral range 400-1000 nm, based on three-channel spectral zonal images and reference spectra measured at several points of the recorded scene. The procedure makes it possible to improve the information content of the data for solving various thematic classification problems.Introduction. Differences in the reflectance spectra of objects is the basis for recognition of such objects by optical remote sensing [1][2][3][4][5]. In studies of the Earth's vegetative cover, in particular forests, the possibilities of remote sensing have been far from completely utilized [6,7]. Most work on remote sensing as applied to forests (as for other vegetation stands) is based on data obtained from space vehicles. In a number of publications, the authors rely on data obtained using spectrometers and video spectrometers mounted on airplanes [8][9][10][11] and also on helicopters [7]. The advantages and disadvantages of airborne imaging from aircraft are briefly discussed in [5,7]. In addition, we note that when aircraft are used, the reliability of the classification of types of stands may be very high. Comparison of the results from using images with different spatial resolution [12] showed that images obtained from low altitudes make it possible to determine forest fragmentation and biodiversity at the level of individual treetops (for inventory purposes), while those obtained from high altitudes make it possible to very accurately classify the composition of the vegetation at the landscape level. One more problem in forest monitoring which is also solved more accurately using aircraft is determination and assessment of damage done by forest fires. This problem is important not only from an economic point of view but also from the standpoint of the effect of fires on global ecology, which is more significant than suggested in [13].An advantage of spectral zonal images obtained on board aircraft is their high spatial resolution, generally matching the rather large dimensions of the frame (capture band); a relative disadvantage is the limited set of spectral zones (bands), which may not always convey the specific spectral details of objects on the underlying surface. However, very often three or four channels are enough to achieve good recognition of vegetation [14]. In this case, in contrast to automatic classification based on the spectral characteristics of the image recorded by a hyperspectral camera, here partially interactive processing is required to ensure high accuracy [15, 16]. The correct choice of the channels optimizes the spectral zonal system with respect to the efficiency/cost criterion and is quite important. This choice depends on the objects to be studied (the underlying surfaces) and what needs to be done. The video spectrometric system VSK-2 that we designed and used for studying forests [5,7] is an imageplane scanner. Each frame records "instantaneously"; in contrast to hyperspectral scanners,...