| Abstract [eng] |
The aim of the project is to analyse continuous tracking of humans within selected method or methods together with related algorithm and use computer program for image processing. Research is done monitoring people and proceeded in MATLAB computer program. In the first part of the project methods and algorithms used in detection and tracking of the objects are analysed. Some of them are described only theoretically, while others are also applied practically. Most of the authors performed experiments, which show not only advantages, but also disadvantages of the described systems, although improvements are being introduced. The majority of algorithms used in tracking of objects are depicted only mathematically, therefore have no practical value. In the second part, all the practically applied methods used for observation and detection of objects are being described. Detection and tracking of objects in MATLAB environment is also presented. Face detection with region of interest used for detection of the object and background subtraction is applied for object tracking. Moreover, methods used in MATLAB environment for picking out features of objects and SVM classificatory parameter, which compares them between each other, are analysed in detail. In the experimental part of the project research is depicted, where it is attempted to apply MATLAB written program for continuous people tracking towards the video clip with multiple people appearing in it. During the experiment with first video, used two approaches for picking out the features of people were employed: LPB and HOG. The best approach for picking out the features was determined. Second video uses more approaches for picking out the features of people. Also, machine learning was done in MATLAB environment together with SVM, which performs comparison of features and identifies if the person was detected before or is newly appeared one. Depending on this data results and conclusions of the experiment are composed. |
