Razvoj sustava za robotsko sortiranje ribe primjenom metoda strojnog učenja

master's thesis

• Title (english): Development of a system for robotic fish sorting using machine learning methods
• Author: Marijo Cepernić
• Mentor: Bojan Jerbić (mentor)
• Mentor: Filip Šuligoj (komentor)
• Granter: University of Zagreb; Faculty of Mechanical Engineering and Naval Architecture; Zagreb
• Defense date and country: 2021-07-14, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Mechanical Engineering

Abstract

Sustav za klasifikaciju i separaciju riba uspješno razvijen u okviru ovog rada predstavlja temelj za snažniju automatizaciju industrije prerade ribe. Industrija u kojoj se, do danas, mahom nalaze linije za preradu s vrlo visokom razinom ljudske radne snage predstavlja ogroman potencijal za snažnijom automatizacijom. Istovremeno, izazovi kod automatizacije linija za preradu riba ograničavaju snažniju zastupljenost ovakvih sustava. Sve intenzivniji ekonomski pritisci na proizvođače ribljih proizvoda predstavljaju svojevrsni okidač u procesu transformacije iz dominantno manualnog proizvodnog sustava na robotizirani.
Industrijski vizijski sustav, umjetna inteligencija i robotski manipulator su tri ključna segmenta sustava priloženog u ovom radu. Uz popratne, ovi segmenti sustava uspješno klasificiraju i separiraju srdele koje se pomiješane s inćunima nalaze na pokretnoj traci. Klasifikacija se izvodi metodama strojnog učenja na temelju podataka dobivenih iz vizijskog sustava. Obrada slika i metode strojnog učenja su implementirane na računalu koji je, s druge strane, povezan s robotskim manipulatorom. Delta robot, koji vrši samu separaciju, je odabran zbog vrlo kratkih vremena ciklusa i ima mogućnost praćenja trake u pokretu. Na temelju informacija koje prima s računala, robot izuzima, orijentira i paletizira jedinke s pokretne trake. Jednom paletizirane, ribe su spremne za daljnju preradu na proizvodnoj liniji.
Maksimalni protok riba kroz razvijeni sustav je relativno velik i to svojstvo, uz činjenicu da se sustav vrlo lako nadograđuje s više robota, predstavlja potvrdu koncepta.

Abstract (english)

The system for classification and separation of a fish successfully developed within this paper is the basis for stronger automation of the fish processing industry. The industry in which, to date, there are mostly processing lines with a very high level of human labor represents a huge potential for more intense automation. At the same time, the challenges of automating fish processing lines limit the stronger presence of such systems. Increasing economic pressures on producers of fish products represent a kind of a trigger in the process of transformation from a predominantly manual production system to a robotic one.
The industrial vision system, artificial intelligence, and the robotic manipulator are the three key segments of the system elaborated in this paper. These system segments successfully classify and separate sardines mixed with anchovies on the conveyor belt. Classification is performed by machine learning methods based on data obtained from the vision system. Image processing and machine learning methods are implemented on a computer which, in turn, is connected to a robotic manipulator. The delta robot, which performs the separation itself, was chosen because of the very short cycle times and has the ability to track the object on the conveyor in motion. Based on the information it receives from the computer, the robot excludes, orients, and palletizes sardines from the conveyor belt. Once palletized, the fish is ready for further processing in the production line.
The maximum flow of fish through the developed system is relatively large and this feature, along with the fact that the system is very easily upgraded with multiple robots, is a proof of the concept.

• Keywords: Klasifikacija riba; Umjetna inteligencija; Delta robot; Vizijski sustav; Klasifikator; Paletizacija; Praćenje trake
• Keywords (english): Fish classification; Artificial intelligence; Delta robot; Vision system; Classifier; Palletizing; Line tracking
• Language: croatian
• URN:NBN: urn:nbn:hr:235:268473
• Study programme: Title: Mechanical Engineering; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka strojarstva (magistar/magistra inženjer/inženjerka strojarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access; Embargo expiration date: 2023-07-01
• Created on: 2021-07-08 12:16:57