| Popis: |
Potreba za detekcijom pokreta i praćenjem objekata pojavljuje se u mnogim primjenama računalnog vida. Detekcija pokreta je proces koji primjećuje promjenu pozicije objekta u odnosu na njegovo okruženje ili promjenu okruženja relativno u odnosu na objekt. Metode detekcije pokreta možemo razvrstati u one jednostavne, koje nisu zahtjevne procesorski ni memorijski, no zato imaju par mana, kao teže detektiranje malih pokreta objekata, kao i manju preciznost i robusnost i one koji koriste kompleksnije algoritme za usporedbu pozadine s trenutnim okvirom. Proučeni su razni načini detekcije pokreta i implementirane dvije metode, dok je jedna metoda testirana na razne promjene praga i različitim uvjetima snimke. S dobro određenim parametrima praga i zamućivanja i pogodnim uvjetima zadana metoda daje jako precizne i robusne rezultate. The need for motion detection and tracking of objects appears in many applications of computer vision. Motion detection is a process that observes change the position of the object in relation to its environment or change environments relative to the object. Methods of motion detection can be divided into those simple, non-demanding methods, but because of that they have a couple of flaws, as it is more difficult to detect small movements of objects, as well as lower accuracy and robustness and those who use more complex algorithms to compare the current frame the background. Various ways of motion detection have been examined and two simple methods have been implemented, while one method was tested on a variety of changes in the threshold and motion blur parameters. With well-defined threshold parameters and motion blur and suitable conditions the default method can give very accurate and robust results. |
