E.A. Samoylin - Dr. Sc. (Eng.), Associate Professor, Professor, MESC «Zhukovsky-Gagarin Air Force Academy» (Voronezh). E-mail: es977@mail.ru M.A. Pantyukhin - Post-graduate Student, MESC «Zhukovsky-Gagarin Air Force Academy» (Voronezh). E-mail: ol-max@mail.ru

Currently, the vast majority of recognition operators used in various optoelectronic systems, focused on the contour representation of an object that allows you to generate invariant features and reduce the dimension of the problem being solved. Meanwhile, the operation of such systems in conditions of laser destructive interference of the opposing side illumination leads to a failure of the photosensitive elements of the optical radiation detectors, in turn, leads to the appearance of pulse noise on the images recorded. The use of well-known gradient algorithms of edge (masks Prewitt, Sobel, Kirsch et al.) In terms of impulse noise on digital images leads to significant errors of the first and second kind, when it detects the contour of the object elements. Accordingly, the purpose is to enhance the accuracy of the gradient separation contour features of objects in digital images recorded by the laser in destructive interference. The idea of the proposed algorithm is to divide the processing procedure in two stages. In the first stage, the image is detected impulse noise, i.e. formed a binary matrix interference assessment. Then, in the second stage adaptation, i.e. changing the weighting values variously masks (Prewitt, Sobel, Kirsch et al.) depending on whether they fall into the failed elements, which in turn indicate individual elements of the interference estimates. These results of numerical studies and examples of image processing show that the proposed adaptive algorithm has fewer total errors of the first and second order compared to the traditional mask and Prewitt masks known adaptation algorithm in the entire range of values of probability of the presence of impulse noise. The algorithm is considered as an example of adaptation weights Prewitt gradient masks, but can easily be generalized to any known gradient masks.

 

1. Djatlov E.I. Mashinnoe zrenie (analiticheskijj obzor) // Matrichnye sistemy i mashiny. 2013. № 2. S. 32−40.
2. Potapov A. Sistemy kompjuternogo zrenija: sovremennye zadachi i metody // ControlEngineering Rossija. 2014. № 1 (49). S. 20−26.
3. Gonsales R., Vuds R.Cifrovaja obrabotka izobrazhenijj. M.: Tekhnosfera. 2005. 1072 s.
4. Kozirackijj JU.L., Glushkov A.N. Sintez lazernykh pomekh koordinato-metricheskim optiko-ehlektronnym sistemam // Radiotekhnika. 2011. № 8. S. 37−41.
5. Samojjlin E.A. Metod adaptacii masok v zadachakh gradientnogo vydelenija konturov izobrazhenijj na fone impulsnykh pomekh // Izvestija RAN. Teorija i sistemy upravlenija. 2010. № 2. S. 118−129.
6. Samojjlin E.A. Sintez statisticheski optimalnykh obnaruzhitelejj impulsnykh pomekh pri filtracii izobrazhenijj // Radiotekhnika. 2009. № 9. S. 25−31.
7. Samojjlin E.A., SHipko V.V., Ivanov M.G. Algoritm obnaruzhenija applikativnykh pomekh na izobrazhenijakh, registriruemykh trekhkanalnym fotopriemnikom v uslovijakh lazernogo podavlenija // Radiotekhnika. 2014. № 9. S. 68−72.