Defining Thin Objects in Computer Vision

Thin image objects detection in computer vision is a challenging task due to the compactness, slenderness, and extremely small coverage of these objects in the images. This makes these objects difficult to detect especially in the presence of background clutter. Efficient detection of thin image objects is of utmost importance in research areas including but not limited to electrical transmission and mooring fault detection systems, medical analysis, collision and obstacle detection in unmanned aerial and autonomous vehicles, plant stem detection in biological studies, road-lane-line detection and so on. Despite the significance of thin objects detection from images, there is no formal and accurate definition of what thin objects in computer vision are. In this research work, we propose and formalize the criteria for defining thin objects in computer vision. We conduct various experiments to identify the parameters and formalize the measures for describing thin objects in several computer vision research applications and other novel scenarios. The proposed research work can be utilized in the future research for the design and development of efficient thin objects detection and localization datasets, algorithms and other relevant tasks.