Autonomous mobile robot programming for human face detection and tracking

U ovom radu, prikazano je programsko rješenje autonomnog mobilnog robota koji detektira lice osobe preko web kamere laptopa, te prati tu osobu na način da ju nastoji držati u sredini ekrana i na određenoj udaljenosti od kamere. U 3. poglavlju je opisana metoda za detekciju lica koja je ostvarena korištenjem Python biblioteke opencv koja ima već gotovi predtrenirani model za detekciju lica i očiju osobe na slici. Robot je programiran u ROS-u na ubuntu operacijskom sustavu pa je 2. poglavlje rada posvećeno ROS-u. Kako bi robot u svakoj iteraciji računao nove iznose linearne i kutne brzine, koristi se neizrazita logika i Mamdanijeva metoda zaključivanja. Na taj način je ostvareno fino gibanje robota bez naglih pokreta i promjena brzine. Neizrazita logika i neizrazito zaključivanje je opisano u 4. poglavlju. U 5. i 6. poglavlju je opisan način na koji se može testirati programski kod za gibanje autonomnog robota, pomoću simulacije u alatu Gazebo. 7. i 8. poglavlje rezervirano je za pokretanje i testiranje stvarnog Huskyja robota koji se nalazi u prostorijama FER-a, a u 9. poglavlju je proveden zaključak rada.In this thesis, a program solution is shown for autonomous mobile robot which detects face of a person using laptop webcam, and follows that person in a way that it keeps certain distance from person all the time. Also, robot is always trying to keep person in center of webcam frame. In 3rd chapter, a model for face detection used in this thesis is introduced. Model is a pretrained face and eye detection model that is part of opencv Python library. Robot is programmed in ROS on ubuntu operational system so 2nd chapter is dedicated to ROS. For robot to make a conclusion about new linear and angular velocity in every iteration, fuzzy logic and Mamdani method of conclusion are used. Because of that, robot never makes sudden moves and keeps its movement smooth which keeps person and robot environment safe. Fuzzy logic and fuzzy conclusion is a topic of 4th chapter. In 5th and 6th chapter, Gazebo simulation for robot movement is described for testing of a program algorithm written in Python. 7th and 8th chapter is reserved for testing and experimenting on real Husky robot that is located in Across laboratory in FER and in 9th chapter a conclusion to thesis is made