Neural network technology adaptation to the small-size objects identification in satellite images of insufficient resolution within the graphic reference images database

Abstract

A novel flowchart for small-size objects identification in satellite images of insufficient resolution within the graphic reference images database using neural network technology based on compromise contradiction, i.e. simultaneously the resolution enhancement of the object segment of input image and the resolution reduction of the reference image to joint resolution through the simulation of the imaging system has been proposed. This is necessary due to a significant discrepancy between the resolutions of the input image and the graphic reference images used for identification. The required level of resolution enhancement for satellite images, as a rule, is unattainable, and a significant coarsening of reference images is undesirable because of identification errors. Therefore, a certain intermediate spatial resolution is used for identification, which, on the one hand, can be obtained, and on the other the loss of information contained in the reference image is still acceptable. The intermediate resolution is determined by simulating the process of image acquisition with satellite imaging system. To facilitate such simulation, it is advisable to perform it in the frequency domain, where the advanced Fourier analysis is available and, as a rule, all the necessary transfer properties of the links of image formation chain are known. Three main functional elements are engaged for identification: an artificial neural network for the resolution enhancement of input images, a module of frequency-domain simulating of the graphical reference satellite imaging and an artificial neural network for comparing the enhanced object segment with the reference model images. The feasibility of the described approach is demonstrated by the example of successful identification of the sea vessel image in the SPOT-7 satellite image. Currently, the works are under way to compare the performance of a neural network platforms variety for small-size objects identification in satellite images aa well as to assess achievable accuracy.

Features of hardware implementation of quasi-continuous observation devices with discrete receivers

“…The search of opportunity to fulfill Eq. ( 6) involves following hardware and algorithmic solutions: (1) constructive improvements [4,5] in using multi-camera observation devices [6] (which, however, complicate the design of observation devices); (2) a decrease in the size of an elementary receiver to the diffraction limit, so more receivers can be placed on a unit of surface matrix [7,8] (the implementation of which is technologically complicated and limited by theoretical limits of applicability); (3) consideration of the design of the real tasks of a receiver for observing objects [9] (it reduces the possibility of observing all objects in the image); (4) taking into account the spectral properties of objects of observation and using additional signs of their recognition [10,11] (during the hyper-spectral observation of an object, a large amount of information is accumulated, and its interpretation requires a high and specific operator qualifications); (5) use of regularly changing (controlled) density of elementary receiversmatrices (RCDOER-matrices) in the structure of the species tool (RCDOER-matrices is only declared, the tasks and conditions of effective work are not formulated); and (6) application of special procedures for processing observation results [12][13][14][15][16] (the possibilities of algorithmic processing are a posterior, and they are always limited to the results of apparatus observation).…”

A method of quasi-continuous image formation in observation devices with discrete receivers

“…There are several known ways to partially solve the problem of limited spatial resolution. They include the following hardware and algorithmic directions: • structural improvements (Popov et al, 2015;Popov, 2018), application of multicamera photographic devices (The largest, 2021) (but they complicate the observation device design); • reduction of elementary receiver size to the diffraction limit while is increasing of number of such receivers in a photosensitive matrix (OWC, 2021; Rehm, 2021) (unfortunately, their realization is technologically difficult and they are limited by theoretical limits of applicability); • taking into account in the design of the receiver tasks for the observation of objects with a finite size of object recognition (Korobchynskyi et al, 2020) (but it is not always possible to predict in advance the required depth of object recognition by its image); • using the features of zonal images of the object and its components (Popov et al, 2007;Ferraris et al, 2018) (at hyperspectral observation of the object is accumulated a huge amount of information, its interpretation requires high and specific qualification of the operator); • application of neural network technologies, algorithms of extrapolation, interpolation, probabilistic analysis and estimation (Kwan, 2018;Stankevich et al, 2020) (it is a very promising areas, which are based on the use of a priori information like the results of observation and documentation of the observation device).…”