COMPUTER BASED PEDESTRIAN COUNTING

V tem diplomskem delu se ukvarjamo z razvojem števca prehodov ljudi, ki temelji na podlagi računalniškega vida. Na začetku predstavimo že obstoječe možne rešitve štetja, zatem pa razvijemo in implementiramo lasten algoritem. V diplomskem delu so opisane vse ključne funkcije potrebne za uspešno prepoznavo objektov na sekvenci slik ter sledenja njihovem gibanju. Z razvito aplikacijo smo dosegli natančnost štetja prehodov nad 80 %.In this thesis we deal with development of pedestrian counting system, based on computer vision. Firstly, exsisting methods of pedestrian counting are reviewed, afterwords we develop and implement our own computare vision based algorithm. In this thesis all the fundamental functions needed for object recognition and object tracking are described. With the developed application we acheaved accuracy of pedestrian counting more than 80 %