Underwater-Drone With Panoramic Camera for Automatic Fish Recognition Based on Deep Learning
Autor: | Shigeru Oyanagi, Takuma Hirayama, Lin Meng |
---|---|
Rok vydání: | 2018 |
Předmět: |
open source hardware
General Computer Science Computer science ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION 02 engineering and technology Field (computer science) 360-degree panoramic image 0202 electrical engineering electronic engineering information engineering General Materials Science Computer vision Underwater business.industry Deep learning Frame (networking) General Engineering underwater-drone deep learning 020206 networking & telecommunications Drone Fish 020201 artificial intelligence & image processing lcsh:Electrical engineering. Electronics. Nuclear engineering Artificial intelligence fish recognition business lcsh:TK1-9971 |
Zdroj: | IEEE Access, Vol 6, Pp 17880-17886 (2018) |
ISSN: | 2169-3536 |
DOI: | 10.1109/access.2018.2820326 |
Popis: | Highly developed drone technology enables the use of drones in a wide variety of areas. However, those drones are mainly used in the unmanned aerial vehicles. We believe that underwater drones will become a big research topic and find a market in the near future. We developed an underwater drone with a 360° panoramic camera acting as the “eye”of the drone. The designs are based on the open-source hardware and will be shared as an open-source for contributing to the innovation of manufacturing including drone. The function of the 360° panoramic camera is generated by correcting the images taken by two fisheye lenses. The underwater drone was designed by extending the Raspberry Pi compute module, the frame was designed by OpenSCAD, and the printed circuit board was designed by MakePro. As for the application of the underwater drone, we focused on fish recognition for investigating fish species in a natural lake to help protect the original environment. Fish recognition is based on deep learning, which is the biggest topic in the artificial intelligence research field today. Experimental results show that the function of the underwater drone achieved at diving in the leak automatically. The 360° panoramic images were generated correctly. Fish recognition achieved 87% accuracy by deep learning. |
Databáze: | OpenAIRE |
Externí odkaz: |