The "dual-objective edge computing single-lens UAV smart vision following technology" developed by Xu Zigui, a professor of smart electrical engineering at the National Pingtung University of Science and Technology, entered the 20th International Invention Exhibition in Warsaw, Poland this month and won the gold medal, and even won the "Best Invention Award at the Conference". Xu Zigui said that through a model with software innovation as the main axis, it is possible to predict the trajectory of the target in the next 10 seconds, the construction cost is reduced by 80%, and even the endurance of the drone can be increased by at least 20%.
The Polish Invention Exhibition is the second largest world invention exhibition in the European Union. This year, it brought together more than 400 innovative works from 31 countries and regions around the world. PSU won the first place with the drone project, which not only brought glory to the school, but also demonstrated to the world Taiwan's high competitiveness in the research and development of smart unmanned vehicles. Xu Zigui pointed out that the team has successfully developed a solution to replace traditional expensive dual-light or triple-light composite photoelectric cameras with low-cost single-lens equipment. By integrating YOLO artificial intelligence, edge computing and dynamic PID flight control technology, the drone has a keen "AI eagle eye" that can autonomously identify, lock and follow moving targets.
"Assisting artificial intelligence with physical models, replacing expensive hardware with software innovation," Xu Zigui said, the most innovative thing is actually using optimized software to lead the main axis of the hardware through the AI model. The team cross-domain introduced the Particle Image Velocimetry (PIV) technology of fluid mechanics and combined it with the explicit method time step prediction method commonly used in computational fluid dynamics to establish a physical prediction model of target motion.
The feedback data is driven by edge computing in real time. By analyzing the target's position, speed and movement direction through continuous images, the system can estimate the possible movement trajectory in the next 10 seconds, reducing the risk of image tracking loss caused by sudden acceleration, turning or temporary occlusion of the target, breaking through the natural dangers of traditional AI image tracking that mainly relies on real-time recognition.
Xu Zigui said that in addition to estimating the trajectory, the system also combines the "area replacement distance" algorithm independently developed by the team, which can accurately estimate the distance and automatically adjust the flight speed. According to the evaluation of the R&D team, this technology can reduce hardware costs by up to 80% and is expected to increase battery life by 20% to 35%, echoing the main axis of "replacing expensive hardware with software innovation".
ScreenTech said that in the future, it will continue to combine the advantages of Taiwan's AI chip, electronic control and drone manufacturing industries, promote the international market layout through technology transfer and industry-university cooperation, and allow the "AI Eagle Eye" developed by ScreenTech to spread its wings from Taiwan and fly to the global technology stage.
