ADAPT: Autonomous Data Acquisition and Processing Technologies Payload for Small Unmanned Aircraft Systems

We propose the continuation of development for the Autonomous Data Acquisition and Processing Technologies (ADAPT) payload. This open source system will manage data collection from onboard sensors, particularly imaging sensors, and provide actionable intelligence by enabling in-flight, real-time application of DNN image processing to support a variety of missions. We will substantially reduce the barrier to training and deploying useful DNN-powered monitoring solutions at the edge by laying the groundwork with with an open-source baseline payload, solving the data collection, geolocation, and model deployment aspects of payload design in a way that is agnostic to the drone platform of choice. For this effort, the particular problem of river ice detection will be pursued in order to demonstrate the efficacy of the ADAPT payload for solving the problem of real time flood monitoring during icing months in Alaska. This will result in a first of its kind dense river ice segmentation dataset, a trained state-of-the-art river ice segmentation model, a small unmanned aerial vehicle (sUAS) payload for real time deployment of the developed model, and open source software and system architecture documentation for ease of reproducibility. ADAPT has the potential to aid a far-reaching customer and user base internationally as it caters to different applications, missions, domains, and industries. Affordability, flexibility, extensibility, accessibility, and the ability to provide real-time intelligent analyses are critical areas where existing systems do not meet this need or are only meeting this need partially. Commercial applications supporting environment monitoring, search and rescue, disaster management and more can be easily realized. Kitware will leverage its proven track record for strategically financing open source solutions to build a coalition of support and an active user community around ADAPT.