izpis_h1_title_alt

Detekcija min na toplotnih slikah
ID Crček, Matevž (Author), ID Skočaj, Danijel (Mentor) More about this mentor... This link opens in a new window

.pdfPDF - Presentation file, Download (15,24 MB)
MD5: FB6E62034ADE05707A4346840131BE7B

Abstract
Diplomska naloga obravnava problem detekcije min na toplotnih slikah. Prikažemo prednosti toplotnih slik v primerjavi z barvnimi pri detekciji min. Analiziramo edino javno dostopno podatkovno množico toplotnih slik min, izpostavimo njene pomanjkljivosti in predlagamo izpeljano podatkovno množico, ki jih odpravlja. Obstoječo metodo za detekcijo min izboljšamo z uporabo YOLOv8, nato pa predlagamo nov pristop k reševanju problema z detekcijo napak na slikah z arhitekturo SegDecNet. Primerjamo uspešnost arhitektur YOLOv5, YOLOv8 in SegDecNet na izvorni in izpeljani podatkovni množici ter pokažemo, da naša metoda dosega boljše rezultate v vseh primerih. Izpostavimo omejitve primerjave rezultatov na izpeljani podatkovni množici ter predlagamo izboljšave podatkovne množice in možne razširitve arhitekture SegDecNet za detekcijo predmetov.

Language:Slovenian
Keywords:računalniški vid, toplotne slike, globoko učenje, konvolucijske nevronske mreže, detekcija predmetov, detekcija napak
Work type:Bachelor thesis/paper
Organization:FRI - Faculty of Computer and Information Science
Year:2024
PID:20.500.12556/RUL-161559 This link opens in a new window
Publication date in RUL:12.09.2024
Views:65
Downloads:19
Metadata:XML RDF-CHPDL DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Secondary language

Language:English
Title:Detection of mines on thermal images
Abstract:
The thesis addresses the problem of mine detection in thermal images. We demonstrate the advantages of thermal images compared to color images for mine detection. We analyze the only publicly available dataset of thermal mine images, highlight its shortcomings, and propose a derived dataset that addresses these issues. We enhance the existing mine detection method using YOLOv8 and then propose a new approach to solving the problem through defect detection using the SegDecNet architecture. We compare the effectiveness of the YOLOv5, YOLOv8, and SegDecNet architectures on the original and derived datasets, showing that our method achieves better results in all cases. We also highlight the limitations of comparing results on the derived dataset and suggest improvements to the dataset and potential extensions of the SegDecNet architecture for object detection.

Keywords:computer vision, thermal images, deep learning, convolutional neural networks, object detection, defect detection

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Back