| Abstract [eng] |
In this Masters's final thesis research on drones usage for person identity recognition in the crowd is performed. The system used for facial recognition in the crowd using drones is developed, implemented and tested. The analytical part analyzes scientific literature on face detection, alignment and recognition algorithms, drone usage for various objects detection. Algorithms use cases for facial recognition in the crowd is analyzed. Existing stationary camera systems using facial recognition technologies are reviewed their shortcomings identified. In the research part of the master thesis system for person identity recognition in the crowd using drones is developed and implemented. Best algorithms and system components for use with video streams from drones camera are selected. Testing on the stationary system is performed, selected algorithms are evaluated with drones work specifics in mind, recognition chain performance is evaluated and optimized. The system used for facial recognition using drones is implemented, it consists of applications for server and web. The developed system is implemented in such way that it can be adapted to changing weather and flight conditions. After the system is implemented it is tested in real usage conditions, optimal flight altitude, distance, lighting conditions and system parameters are identified. Person identity recognition test in the crowd is performed. Structure: introduction, facial recognition methods and systems, recognition system development and component selection, stationary system testing, system implementation and testing, conclusions, references. Thesis consist of: 82 p. text without appendices, 37 pictures, 1 tables, 63 bibliographical entries. |
