Image distortion correction and stitching are well-established problems in computer vision, typically addressed using traditional methods that rely on calibration targets or feature extraction from the image. However, these conventional approaches face limitations when applied to large-scale, textureless environments, such as ice rinks, where calibration targets are impractical, and stitching is rendered unreliable due to the lack of texture clues. This research introduces a novel method that uses spiral codes mounted on a mobile robot and ultrasound sensors to effectively rectify image distortion and provide clues for image stitching in such challenging settings.

Our approach is particularly beneficial in environments lacking discernible features or textures, where traditional image stitching methods fail. Leveraging spiral patterns to establish correspondences between overlapping image areas enables accurate stitching even in featureless scenes. Integrating an autonomous mobile robot with ultrasound sensors further provides ground truth measurements for distortion correction and image stitching.

The thesis first formulates the research questions with appropriate motivation and justification, then introduces the suggested solution, and ultimately demonstrates the combined application of distortion correction and image stitching using the proposed methodology.