prikaz prve stranice dokumenta Detekcija vozila u okruženju autonomnog vozila u svrhu upozorenja na potencijalnu koliziju
  Download
PDF 4.52 MB
master's thesis
Detekcija vozila u okruženju autonomnog vozila u svrhu upozorenja na potencijalnu koliziju
Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2019. urn:nbn:hr:200:192117
Gluhaković, Mario
Josip Juraj Strossmayer University of Osijek
Faculty of Electrical Engineering, Computer Science and Information Technology Osijek
Department of Communications
Chair of Electronics and Microelectronics

Institutional repository: Faculty of Electrical Engineering, Computer Science and Information Technology Osijek

Cite this document

Gluhaković, M. (2019). Detekcija vozila u okruženju autonomnog vozila u svrhu upozorenja na potencijalnu koliziju (Master's thesis). Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek. Retrieved from https://urn.nsk.hr/urn:nbn:hr:200:192117

Gluhaković, Mario. "Detekcija vozila u okruženju autonomnog vozila u svrhu upozorenja na potencijalnu koliziju." Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2019. https://urn.nsk.hr/urn:nbn:hr:200:192117

Gluhaković, Mario. "Detekcija vozila u okruženju autonomnog vozila u svrhu upozorenja na potencijalnu koliziju." Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, 2019. https://urn.nsk.hr/urn:nbn:hr:200:192117

Gluhaković, M. (2019). 'Detekcija vozila u okruženju autonomnog vozila u svrhu upozorenja na potencijalnu koliziju', Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, accessed 19 May 2024, https://urn.nsk.hr/urn:nbn:hr:200:192117

Gluhaković M. Detekcija vozila u okruženju autonomnog vozila u svrhu upozorenja na potencijalnu koliziju [Master's thesis]. Osijek: Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek; 2019 [cited 2024 May 19] Available at: https://urn.nsk.hr/urn:nbn:hr:200:192117

M. Gluhaković, "Detekcija vozila u okruženju autonomnog vozila u svrhu upozorenja na potencijalnu koliziju", Master's thesis, Josip Juraj Strossmayer University of Osijek, Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, Osijek, 2019. Available at: https://urn.nsk.hr/urn:nbn:hr:200:192117

Metadata
TitleDetekcija vozila u okruženju autonomnog vozila u svrhu upozorenja na potencijalnu koliziju
Title (english)Vehicles detection in the autonomous vehicle environment for the purpose of potential collision warning
AuthorMario Gluhaković
MentorMarijan Herceg (mentor)
Committee memberMarijan Herceg (predsjednik povjerenstva)
Committee memberMario Vranješ (član povjerenstva)
Committee memberRatko Grbić (član povjerenstva)
GranterJosip Juraj Strossmayer University of Osijek
Faculty of Electrical Engineering, Computer Science and Information Technology Osijek
(Department of Communications)
(Chair of Electronics and Microelectronics)
Osijek
Defense date and country2019-09-17, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Artificial Intelligence
Abstract
Tema ovog diplomskog rada je „Detekcija vozila u okruženju autonomnog vozila u svrhu upozorenja na potencijalnu koliziju“. Jedna od važnih stvari u autonomnoj vožnji je percepcija okoline vozila. U ovom radu razvijen je sustav koji detektira vozila u okruženju autonomnog vozila i obavlja procjenu udaljenosti detektiranih vozila. Za detekciju vozila koristi se Yolo v2 algoritam koji je istreniran na novo kreiranom skupu slika, te obavlja detekciju četiri klase vozila: automobil, kombi, kamion i autobus. Sustav je implementiran u ROS pomoću čvorova. Za procjenu udaljenosti kreirana su dva čvora. Jedan čvor se koristi za procjenu udaljenosti u Carla simulatoru i dobiven je eksperimentalnim putem, dok se drugi čvor koristi za procjenu udaljenosti u stvarnom svijetu i zasnovan je na procjeni udaljenosti pomoću parametara kamere i referentne širine vozila u stvarnom svijetu. Obavljeno je testiranje algoritma na stolnom računalu, korištenjem slika iz simulatora i stvarnog svijeta. Uočeno je, da algoritam radi na prosječno 40 okvira u sekundi na Nvidia GTX1060 (3Gb) grafičkoj kartici, što znači da algoritam radi u stvarnom vremenu. Razvijeni algoritam omogućava detekciju objekata i kada se samo dio objekta nalazi unutar slike ili kada se više objekata preklapa. Moguća je detekcija vozila iz svih pozicija oko vozila. Prilikom testiranja je uočeno, da algoritam ne detektira objekte na udaljenosti većoj od 50 metara, te da slabije detektira objekte po noći i kiši. Također je uočeno, da se povećavanjem udaljenosti preciznost procijenjene udaljenosti smanjuje, te da je potrebno obaviti određene dorade da bi sustav imao bolju detekciju i preciznost procjene udaljenosti. Razvijeni algoritam je lako prilagodljiv, te ga je lako nadograditi da detektira i druge objekte u prometu.
Abstract (english)
The topic of this masterwork is ''Detection of vehicles in an autonomous vehicle environment for the purpose of warning of potential collision''. One of the important things in an autonomous drive is the perception of the vehicle's environment. In this paper is developing a system that detects vehicles in an autonomous vehicle environment and carries out an assessment of the distance of detected vehicles. For vehicle detection, a Yolo v2 algorithm is used, which is trained on a newly created image set and carries out detection of four classes of vehicle: car, van, truck and bus. The system is implemented in ROS using nodes. Two nodes were created to estimate the distance. One node is used to estimate the distance in the Carla simulator and is obtained experimentally, while the second node is used to estimate the distance in the real world and is based on distance estimation using the camera parameters and the reference width of the vehicle in the real world. The algorithm testing on the desktop computer was done using images from the simulator and the real world. It is noted that the algorithm runs at an average of 40 frames per second on the NVIDIA GTX1060 graphics card, which means that the algorithm works in real time. The developed algorithm allows detection of objects, even when only part of the object is inside the image or when more than one object overlaps. Vehicle detection from all positions around the vehicle is possible. It was noticed that the algorithm does not detect objects at a distance more than 50 meters, and that it detects objects at night and rain less often. It has also been noted that by increasing the distance, the precision of the estimated distance is reduced and certain adjustments have to be made to make the system better detection and precision for the distance estimation. The developed algorithm is easily customizable and easy to upgrade to detect other objects in traffic.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:200:192117
Study programmeTitle: Graduate University Study Programme in Electrical Engineering; specializations in: Communications and Informatics, Power Engineering
Course: Communications and Informatics
Study programme type: university
Study level: graduate
Academic / professional title: magistar/magistra inženjer/inženjerka elektrotehnike (magistar/magistra inženjer/inženjerka elektrotehnike)
Type of resourceText
File originBorn digital
Access conditionsOpen access
Terms of useLicence In copyright icon
Created on2019-12-23 11:37:02