Researchers develop a quality design method for real-time videos from uncrewed aerial vehicles

Researchers at the University of Tsukuba have developed the SPADE (Simulator-assisted PerformAbility Design methodology for UAV-based Systems) approach for accurate quality assessments of uncrewed aerial vehicle (UAV) video surveillance systems. The work is published in the journal Future Generation Computer Systems.

UAV video surveillance systems equipped with computers can monitor real-time images of areas that are not easily accessible to humans. However, a high-quality, reliable video surveillance system that properly integrates computers, UAVs, networks, and surveillance software is difficult to design.

During the early design phase, before the system is implemented on-site, evaluating the quality and considering uncertain factors that might affect the system’s performance are complicated tasks.

SPADE creates a virtual environment closely resembling the actual operating environment of a UAV video surveillance system. Using video data of a UAV flying in this virtual environment, SPADE enables detailed quality evaluations without real-world manipulations.

SPADE collects virtual images using a UAV flight simulator. The recorded images are processed by an actual computer, which measures the processing accuracy, performance, and power consumption of the UAV.

These data are introduced to a state transition model of the system for quality evaluation and trade-off analysis. The SPADE method can evaluate the effects of resolution differences among the input images on the quality of the system for real-time object detection by a UAV.

Various UAV video surveillance systems can be designed through this approach. Future research will develop methods that efficiently design complex, large-scale systems involving multiple UAVs working in coordination.