University of Twente Student Theses
Object Detection in Low Resolution Long Wavelength Infrared Images in Maritime Environment
Verburg, F.M. (2024) Object Detection in Low Resolution Long Wavelength Infrared Images in Maritime Environment.
Full text not available from this repository.
| Full Text Status: | Access to this publication is restricted |
| Embargo date: | 30 June 2025 |
| Abstract: | This research addresses the challenge of object detection in maritime environments using long-wavelength infrared (LWIR) images, a critical task for autonomous sailing. While significant progress has been made in object detection for the automotive industry, resulting in numerous datasets and benchmarks, the maritime domain lacks similar research and resources. The maritime environment differs a lot from the automotive environment. Cars have lights to illuminate the roads in the dark and vessels do not. This makes object detection in infrared images a crucial task for autonomous sailing. The objective is to determine whether inexpensive sensors that produce low-resolution images, requiring minimal processing power, are sufficient for effectively performing this task. To enable this research we created a low-resolution LWIR maritime dataset in the inland area with approximately 5900 images, 6700 vessel labels and 320 buoy labels. Three state-of-the-art object detection models –YOLOv11, Faster R-CNN, and YOLO-FIRI– are evaluated on this dataset to check their ability to detect objects in low-resolution infrared images in maritime environment. Each model is trained on the raw, the colour inverted and the colour inverted + histogram equalized dataset. The results show that the best performing model in terms of recall is Faster R-CNN. The model with the highest mAP@0.5 is YOLOv11 in combination with inversion. |
| Item Type: | Essay (Master) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Embedded Systems MSc (60331) |
| Link to this item: | https://purl.utwente.nl/essays/104697 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page