ChangeAerial Awarded SBIR from the National Science Foundation

July 2024 – ChangeAerial was awarded a Small Business Innovative Research (SBIR) Phase-1 grant from the National Science Foundation (NSF). This grant is designed to fund high-risk, high reward research, adding to ChangeAerial’s already novel and innovative N-SUITE software tools. The project title is “Integrating deep learning algorithms for UAS-based infrastructure inspection: Path to fully automated, commercially viable and scalable monitoring,” and it was funded by NSF’s Translational Impacts (TI) division, within the Directorate for Technology, Innovation and Partnership (TIP).

Existing AI algorithms for detecting defects during electric utility and other inspections are designed to detect limited numbers of specific, pre-identified types of damage. These AI algorithms are trained using hundreds or thousands of examples of particular types of defect. While these algorithms detect and identify the specific problems with roughly 80-90% accuracy, they provide 0% detection accuracy for the infinite number of potential defects that were not trained on (as well as variations based on magnitude of the problem, illumination conditions, material types, and scene backgrounds that vary by geography, etc.). ChangeAerial’s long-term monitoring approach based on automated, pinpoint imaging and detailed change detection is ideal for detecting any variety of defect-related change in appearance. However, while defects are detected accurately, they are not identified. During the SBIR Phase 1 project, ChangeAerial will combine the best of both worlds into a single software application. This will include detailed defect detection based on change monitoring (with ChangeAerial’s proprietary AI-based algorithm), and the capability to label the defect using more traditional CNN-based AI algorithms. In addition, defects detected using ChangeAerial’s novel AI-based change detection can be used to further train CNN algorithms (increasing the variety of defects that can be identified), and change detection can also be used to categorize the severity of change over time and prioritize repairs of detected defects.

Our overall innovation is an integration of three advanced imaging technologies: (1) repetitive, location-based multi-temporal imaging approach called Repeat Station Imaging (RSI © ), (2) UAS capable of navigating and triggering cameras with decimeter positional accuracy using survey-grade global navigation satellite systems (GNSS), and (3) a novel AI-based defect detection approach to automatically and accurately detect even ultra-fine-scale damage on infrastructure apparatus at scale using change detection instead of manual image interpretation or limited AI-based analyses of single-date imagery. Together, these technologies enable a future where UAS imaging, data handling, and image analysis are automated and provide efficient and effective monitoring of electric utility towers, bridges, and other infrastructure that must be maintained to avoid catastrophic disasters such as wildfire ignition and bridge collapse. Using our innovation, UAS images with millimeter (mm) spatial resolution captured from nearly the same location in the sky over time are very accurately co-aligned, enabling subtle damage to be automatically detected with AI.