- Reclamation
- R&D
- Research Projects
- Improving UAS-derived photogrammetric data and analysis accuracy and confidence for high-resolution data sets using artificial intelligence and machine learning
Improving UAS-derived photogrammetric data and analysis accuracy and confidence for high-resolution data sets using artificial intelligence and machine learning
Project ID: 20105
Principal Investigator: Matthew Klein
Research Topic: Condition Assessment
Funded Fiscal Years:
2020
Keywords: None
Research Question
Photogrammetry has been used at Reclamation for over a decade and the technology is becoming more robust and widely accepted especially with the explosion of the UAS market. As with any new emerging technology, the applications for photogrammetric 3D modeling and its products grow every day. However, given recent issues with the derived geospatial products, it seems apparent that Reclamation develop a procedure to standardize photogrammetric data products to allow for the highest accuracy between models processed at different times for change detection. In addition, with increasing understanding and reliability of artificial intelligence and machine learning, there seems to be untapped potential for improved processing results and reduced labor related to processing photogrammetric derived data products.
Given that georeferenced data products are only as accurate as the surveying equipment, errors between 1-2 inches can exist in the georeferencing markers making comparisons between products at less than the georeferenced resolution dangerous. This is a problem since Reclamation's high-resolution photogrammetric products are often accurate to 0.12 inches making comparisons that can capture movement less than 1-2 inches difficult. Methodology will be investigated with input from Reclamation's Professional Land Surveyors to improve change detection between models captured at different times or under different conditions.
In addition, processing UAS photogrammetric data creates a large amount for processing and with the prevalence of artificial intelligence routines and machine learning, it seems natural to investigate leveraging the technology to assist the manual analysis process. For example, once a high resolution orthophoto of concrete deterioration on a dam face is completed, the engineer traces over the existing cracks using a CAD software to number and measure the cracks and crack lengths. This is time consuming and can be prone to errors and is subject t
Need and Benefit
UAS derived photogrammetric products contain a large amount of potential information that can be less accurate than required for analysis and time consuming to analyze manually. By formulating a standard reference protocol and applying machine learning/artificial intelligence, this information will be unlocked to provide detailed analysis of Reclamation's assets for better informed decision making.
Contributing Partners
Contact the Principal Investigator for information about partners.
Research Products
Bureau of Reclamation Review
The following documents were reviewed by experts in fields relating to this project's study and findings. The results were determined to be achieved using valid means.
Identifying Cracks in Concrete from Previously Collected UAS Data Using Deep Learning (final, PDF, 4.1MB)
By Matthew Klein, Zach Leady
Report completed on September 30, 2020
Automatic Concrete Crack Detection Using Previously Collected UAS Data (final, PDF, 348KB)
By Mathew Klein
R&D Bulletin completed on March 30, 2021
