Date of Award
Fall 12-13-2024
Degree Type
Dissertation
Degree Name
Doctor of Philosophy - Forestry
Department
Forestry
First Advisor
Dr. Yanli Zhang
Second Advisor
Dr. I-Kuai Hung
Third Advisor
Dr. David Kulhavy
Fourth Advisor
Dr. Daniel R. Unger
Abstract
Drones can now be used to quickly collect imagery data in a highly automated way; however, individual images must be combined to form an orthomosaic or 3-Dimentional (3D) model using photogrammetry software. Currently, the existing software may generate erroneous output in the form of artifacts or positional errors caused by homogeneous areas, light reflections, object movement between photos, or sub-optimal algorithms. The goal of this research was to develop preprocessing algorithms that would filter movement (or other time or position-based differences) and areas of homogeneity. The hypothesis is that filtering these parts of the image would reduce artifacts and improve the positional accuracy of the resulting orthomosaic images and 3D models. Software improvements to reduce these errors will be especially useful if delivered in an open-source product. Python code was developed to preprocess the images that were input to OpenDroneMap (ODM). The system was tested on a variety of different datasets that each contained a subset of the characteristics that often cause problems (movement, reflection, or undifferentiated areas). Various combinations of filters (treatments) were applied to the datasets and the 2D and 3D results were reviewed for a reduction in artifacts. The results were significantly better with respect to artifacts, but no significant improvement in positional accuracy was observed except in the cases where the drone stopped when capturing an image.
Repository Citation
Ironsmith, Eddie, "UNMANNED AERIAL SYSTEMS (UAS) IMAGE PREPROCESSING TO REDUCE ARTIFACTS AND IMPROVE GEOMETRIC REGISTRATION WHEN GENERATING ORTHOPHOTO MOSAICS AND 3D MODELS" (2024). Electronic Theses and Dissertations. 582.
https://scholarworks.sfasu.edu/etds/582
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
