Camera-based monitoring systems have enormous—yet currently untapped—potential to collect continuous observations for water infrastructure and hydrologic monitoring by extracting measurements from imagery and video, such as surface velocity and depth. Noncontact optical measurement systems can complement and, in some cases, replace conventional sensing technology. While computer vision methods can provide promising and accurate results, practical barriers exist to the widespread implementation and operationalization of optical stream gauging. Common challenges include variable environmental conditions, image and video data management, and systems integration. Deep learning methods have emerged as a promising method to segment water from other objects in images under varying conditions, thus making optical gauging methods more generalizable. Further, cloud computing technologies can be implemented for real-time and on-demand image processing to extract data from images. This talk will demonstrate how deep learning and cloud computing can be deployed with low-cost RGB imaging systems in the field for monitoring water infrastructure, and discuss some of the operational challenges associated with deploying these systems.
.jpg "Sierra Young (she/her/hers) photo")
