a Study of Segmentation Techniques for High-Resolution Uav Imagery

Image segmentation is an established image processing strategy finding its application in different domain of research and implementation like computer vision, photogrammetry, medical imaging, etc. Various general-purpose techniques have been developed for image segmentation. The current study review and compare some of the most frequently used image segmentation technique, namely Image thresholding, Edge-based segmentation, Region-based segmentation, Image clustering, Watershed based segmentation, and ANN-based segmentation. The quality of the individual segmentation results is evaluated based on very high-resolution orthoimage developed with UAV acquired overlapped images for the selected study area covering urban and vegetation features. This is accomplished by qualitative comparison using human visual indexing, which is supplemented by a detailed quantitative analysis encompasses different statistical parameters. The results are assessed and discussed for different land features separately.