Application of computer vision in fish intelligent feeding system—A review

Tracking and Analysis of the Movement Behavior of European Seabass (Dicentrarchus labrax) in Aquaculture Systems

“…can collect behavioral data such as the location of fish in three dimensions, swimming trajectory, acceleration, pressure or muscular activity, swimming depth, body temperature, and acceleration (Pittman et al, 2014;Føre et al, 2017;Macaulay et al, 2021b). Moreover, progress is being made in the development of automatic methodologies that collect and analyze data from a wide range of camera systems, e.g., single or stereo cameras that can exploit different spectra of light such as the visible or infrared (Joseph et al, 2017;Saberioon et al, 2017;An et al, 2021;Zhao et al, 2021). Intelligent monitoring and control methods using mathematical models and computer vision have been developed as a result (Killen et al, 2017;Føre et al, 2018;Zhou et al, 2018;Awalludin et al, 2020;An et al, 2021;Zhao et al, 2021).…”

“…Camera-based methodologies for aquatic monitoring have been drawing increasing attention as they enable a non-extractive and non-lethal approach to studying fish behavior (Papadakis et al, 2012;Liu et al, 2014;Wei et al, 2014;Pautsina et al, 2015;Zhou et al, 2017;Måløy et al, 2019;Pawar and Attar, 2019;Li et al, 2020a;An et al, 2021;Barreiros et al, 2021;Yang et al, 2021) and could be vital for the development of intelligent control methods or systems with efficient monitoring and decision-feedback tools. For example, infrared imaging has been used to track fish and study feeding behavior (Pautsina et al, 2015;Zhou et al, 2017).…”

“…Qian et al (2016) used a single camera system and computer vision techniques to track individual fish movement in tanks. A thorough presentation of methodologies can be found in the reviews of Zhou et al (2018), An et al (2021), andYang et al (2021). Although these methodologies are useful, they usually require specific technological equipment or are species-specific so they lack widerange applicability.…”

“…Individual-level movement analysis can be achieved following a three-step process (Li et al, 2020a;An et al, 2021;Zhao et al, 2021): the detection of the fish, the association of the positions of the detected fish between consecutive frames (fish tracking), and the extraction of swimming feature parameters such as speed, direction, etc. Fish detection is challenging because of the inconsistent illumination in the aquaculture systems, the low contrast between the fish and the background, the low signal-to-noise ratio, the image deformation that results from the scattering and absorption of different wavelengths of light due to particles in the water, the frequent occlusions, and the dynamic background of the aquatic environments (Spampinato et al, 2008).…”

“…Fish detection is challenging because of the inconsistent illumination in the aquaculture systems, the low contrast between the fish and the background, the low signal-to-noise ratio, the image deformation that results from the scattering and absorption of different wavelengths of light due to particles in the water, the frequent occlusions, and the dynamic background of the aquatic environments (Spampinato et al, 2008). Challenges in the fish association step include the partial or total occlusion of the fish by other individuals, and the appearance and disappearance in the image of individuals (Zhou et al, 2018;An et al, 2021;Yang et al, 2021).…”