Praćenje više gibajućih objekata pomoću mješavina Gaussovih razdioba

master's thesis

• Title (english): EN: Multi-object Tracking via Gaussian Distribution Mixtures
• Author: Nikola Arambašić
• Mentor: Ivan Marković (mentor)
• Committee member: Ivan Petrović (član povjerenstva)
• Committee member: Marija Seder (član povjerenstva)
• Committee member: Juraj Peršić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2022-07-12, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Electrical Engineering

Abstract

Praćenje više gibajućih objekata složen je problem čiji je cilj procijeniti broj objekata u određenom području zajedno s kinematičkim stanjima svakog objekta. Uz nesigurnost mjernog i procesnog modela iz klasične Bayesovske estimacije, postoji mnogo dodatnih izvora nesigurnosti u praćenju više objekata. Na primjer, broj objekata se može mijenjati s vremenom, a neki objekti mogu biti neotkriveni u nekim okvirima. Obično ne postoji ni način da se odredi koje detekcije odgovaraju kojem objektu, stoga je u takvim problemima potrebna metoda povezivanja podataka. Korištenjem teorije slučajnih konačnih skupova moguće je proširiti Bayesovu rekurziju na slučaj s više objekata. Međutim, ne postoje opća rješenja zatvorenog oblika. Ipak, postoje mnoge aproksimacije Bayesove rekurzije s više objekata temeljene na Gaussovim mješavinama. Cilj ovog diplomskog rada je implementirati algoritam praćenja više objekata Gaussove mješavine u MATLAB-u i testirati ga u simulacijama.

Abstract (english)

Multi-object tracking is a complex problem where the goal is to estimate the number of objects in the surveillance region together with the kinematic states of each object. In addition to the measurement and process model uncertainty from classical estimation, there are many additional sources of uncertainty in multi-object tracking. For example, the number of objects may change with time and some objects might be undetected in some frames. Usually, there is also no way to determine which detections correspond to which object, hence a data association method is required in such problems. By using random finite sets theory, it is possible to extend the Bayesian recursion to the multi-object case. However, there exists no general closed-form solutions. Nonetheless, there are many approximations to the multi-object Bayesian recursion based on Gaussian mixtures. The goal of this thesis is to implement a Gaussian mixture multi-object tracking algorithm in MATLAB and test it in simulations.

• Keywords: Praćenje više gibajućih objekata; Bayesova estimacija stanja; slučajni konačni skupovi; PHD filter; smanjenje Gaussovih komponenti
• Keywords (english): Multi-object tracking; Bayesian state estimation; Random Finite Sets; PHD filter; Gaussian component reduction
• Language: english
• URN:NBN: urn:nbn:hr:168:467484
• Study programme: Title: Electrical Engineering and Information Technology; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka elektrotehnike i informacijske tehnologije (magistar/magistra inženjer/inženjerka elektrotehnike i informacijske tehnologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 11:41:32