An investigation of image processing techniques for substrate classification based on dominant grain size using RGB images from UAV

摘要

Imagery collected with an unmanned aerial vehicle (UAV) in conjunction with image processing provides new sources of environmental intelligence data and can be implemented in river habitat studies. High-resolution RGB orthomosaic images with 1 cm/px resolution are generated from RGB images acquired with a UAV. Ground truth mapping of the dominant substrate of the river bottom is then used to classify each spatial region. Several texture parameters are examined using image processing techniques to determine the presence and extent of each of the dominant grain classes, providing a method to classify and map the river bed. The method differentiates between submerged, dry exposed, and vegetated regions. The image cover was classified via application and examination of a variety of pixel-based image classification methods. The highest classification accuracy for pixel based analysis was achieved using the thresholding and masking algorithm which achieved an overall 97% correct classification. In addition, object-based image classification was applied using different grey-level co-occurrence matrices (GLCM) in all directions. The classification accuracy for segmentation based classification was found to be lower, at 61%.