Many recent studies have focused on developing image reconstruction algorithms in optical systems based on flat optics. These studies demonstrate the feasibility of applying a combination of flat optics and the reconstruction algorithms in real vision systems. However, additional causes of quality loss have been encountered in the development of such systems. This study investigates the influence on the reconstructed image quality of such factors as limitations of mass production technology for diffractive optics, lossy video stream compression artifacts, and specificities of a neural network approach to image reconstruction. The paper offers an end-to-end deep learning-based image reconstruction framework to compensate for the additional factors of quality losing. It provides the image reconstruction quality sufficient for applied vision systems.

СВЕДЕНИЯ ОБ АВТОРАХКосьянчук Владислав Викторович, доктор технических наук, профессор, первый заместитель генерального директора ФГУП «Государственный научно-исследовательский институт авиационных систем»,

The aircraft integrated control system reconfiguration laws under failures of actuators, calculated disregarding physical constraints on control surfaces saturation, can lead to a complete loss of aircraft controllability and stability. Despite the large number of scientific publications in this field, practical systematic results have been obtained only for SISO (single input – single output) systems. Problems of the convergence of iterative algorithms restricting the set of admissible solutions and the conservatism of the reconfiguration laws designed using weight matrices do not allow solving this problem in general. For complex MIMO (multi input – multi output) systems there is still no widely accepted universal approach. In this work, control surfaces constraints are regarded in terms of the power of reconfiguration control. It is shown that by slight modification of pseudoinverse (optimal) solution it is possible to obtain approximate pseudoinverse (suboptimal) solutions with priory known minimum power (compensation matrix norm) and error (residual matrix norm) of the reconfiguration for a given degree of approximation. This allows for a multistep consistent reduction in power and increasing in error of reconfiguration, until an acceptable solution is obtained. By providing the greater reconfiguration error at each step we have additional freedom in reducing the reconfiguration power. This leads to a decrease in the amplitude of the deviations of the control surfaces, to which the signals from the failed control channels are redistributed. The simulation example of the aircraft integrated control system reconfiguration under the stabilizer’s actuator failure is presented. It is shown that the pseudoinverse reconfiguration problem solution leads to the significant ailerons’ constraints violation and the loss of aircraft controllability. Regarding control constraints solution reduces several times the deviation of the control surfaces and provides an effective problem solution in the permissible power and error reconfiguration range.

No abstract