| Popis: |
V diplomskem delu najprej opišemo osnovne principe delovanja kamer in podrobneje opišemo sistem leč. V poglavju o metodah za avtomatsko fokusiranje pregledamo načine avtomatskega fokusiranja, pri čemer se podrobneje osredotočimo na metode za izračun kontrasta. Opisano teoretično znanje uporabimo pri izdelavi strojno-programskega kompleksa za avtomatsko fokusiranje kamer, ki predstavlja jedro našega diplomskega dela. Natančno predstavimo zgradbo in funkcionalnost strojnega dela naše rešitve, medtem ko pri programskem delu naše rešitve opišemo ne le njegovo funkcionalnost in implementacijske detajle, ampak tudi razloge za razvoj takšnih metod. Skupno smo razvili in implementirali pet različnih metod za avtomatsko fokusiranje kamer, pri čemer tri metode temeljijo izključno na informaciji o kontrastu v opazovanem območju slike, preostali dve metodi pa informacijo o kontrastu kombinirata z na novo vpeljano metriko za ocenjevanje ostrine v sliki. Vse implementirane metode smo tudi preizkusili, tako kvalitativno kot kvantitativno. Pripravili smo deset različnih scen, na katerih smo izvajali teste. Za posamezni eksperiment smo izračunali napako pri fokusiranju. Rezultate smo zbrali in izračunali povprečno napako in standardni odklon za izbrano metodo fokusiranja kamer. Rezultati so pokazali, da je najboljša metoda fokusiranja tista, ki temelji na kontrastu, izračunanem po metodi RMS (angl. root mean square). First, the basic principles of camera operation and lens system are described in detail. In the chapter on methods for automatic focusing the different ways of automatic focusing are reviewed, with further stress on the methods for calculating the contrast. We use the theoretical knowledge to create an application for automatic focusing, which is the core of the diploma thesis. We present the structure and the functionality of the hardware part of the solution, and on the software part we present not only its functionality and implementation details, but also the reasons for developing such methods. We have developed and implemented five different methods for automatic focusing, with three methods based solely on information about the contrast in the observed image area, and the remaining two methods combine the information on contrast with newly introduced metrics to evaluate the sharpness of the picture. All the implemented methods have been tested both in qualitative and quantitative terms. We prepared ten different scenes, in which we performed tests and calculated the error at focusing. We reviewed the results and calculated the average error and standard deviation for each method for focusing. The results have shown, that the best method is the one based on the contrast that was calculated using the RMS (root mean square) method. |
