跟踪(教育)
计算机科学
计算机视觉
跟踪系统
人工智能
作者
Aliaksandr Pautsina,Petr Císař,Dalibor Štys,Bendik Fyhn Terjesen,Åsa Maria O. Espmark
标识
DOI:10.1016/j.aquaeng.2015.09.002
摘要
Abstract Many fish rearing infrastructures are already equipped with human-operated camera systems for fish behavior monitoring, e.g. for stopping the feeding system when the fish is satiated or for monitoring of fish behavior abnormalities caused by poor water quality or diseases. The novel infrared reflection (IREF) system for indoor 3D tracking of fish demonstrated in the current study allows for automation of fish behavior monitoring, reducing the system running costs by eliminating the need for continuous human monitoring and increasing the behavioral analysis accuracy by excluding the human subjectivity factor. The operating principle of this system is based on the effect of strong absorption of near infrared (NIR) range light by water, thus allowing estimation of fish distance based on the corresponding fish object brightness on the camera image. The use of NIR illuminator as a part of the IREF system allows fish behavior monitoring in the dark so as not to affect fish circadian rhythm. A system evaluation under aquaculture facility conditions with Atlantic salmon ( Salmo salar ) using flow-through water in tanks, showed the mean depth estimation error was equal to 5.3 ± 4.0 (SD) cm. The physiological variations among conspecific individual fish introduced the mean depth estimation error of 1.6 ± 1.3 (SD) cm. The advantages of the IREF system over well-known stereo vision systems are lower hardware cost and less computationally intensive 3D coordinates estimation algorithm, while the disadvantage is lower accuracy that is nevertheless acceptable for most applications of aquaculture fish monitoring.
科研通智能强力驱动
Strongly Powered by AbleSci AI