// projects / #5
Active
Modeling and prototyping of technological processes of video analytics in the robotic maintenance of industrial aquaculture
// properties
- Period
- 2026 — 2028
- Project lead
- Ronzhin A.L.
- Grant No.
- 26-19-00326
// description
In the course of fundamental research, methodological, algorithmic, and hardware-software support for a prototype video analytics software system for aquaculture in robotic fish maintenance in cages and closed-circuit water supply installations will be developed, including the following main results with scientific novelty:
- An expert-annotated dataset of fish images, characterized by representative data on the physiological state and behavior of individuals and groups of fish in cages and closed-circuit water supply installations, supplemented with synthesized data with several types of augmentation: rotation, specular reflection, brightness, contrast, scaling, and providing training for neural models to solve the problem of automated detection of fish diseases, feeding intensity management.
- Algorithms, structural, functional and neural models, architecture and prototype of a video analysis software system for aquaculture, characterized by the integration of streams of input data from the aquatic environment, images from video cameras and software modules for pre-processing images of underwater filming, fish detection, tracking the movement of a group of fish, analyzing fish behavior, providing prediction of fish diseases, automation of feeding and others technological processes of robotic maintenance of industrial aquaculture.
- Technical specification for R&D on the development of a video analytics software and hardware complex in the robotic maintenance of industrial aquaculture, featuring a description of options for installing multimedia recording equipment on the physiological state and behavior of fish, activation tools responsible for the technological processes of fish farming operations, taking into account the integration of image processing software modules in predicting fish diseases and automating feeding.
- An expert-annotated dataset of fish images, characterized by representative data on the physiological state and behavior of individuals and groups of fish in cages and closed-circuit water supply installations, supplemented with synthesized data with several types of augmentation: rotation, specular reflection, brightness, contrast, scaling, and providing training for neural models to solve the problem of automated detection of fish diseases, feeding intensity management.
- Algorithms, structural, functional and neural models, architecture and prototype of a video analysis software system for aquaculture, characterized by the integration of streams of input data from the aquatic environment, images from video cameras and software modules for pre-processing images of underwater filming, fish detection, tracking the movement of a group of fish, analyzing fish behavior, providing prediction of fish diseases, automation of feeding and others technological processes of robotic maintenance of industrial aquaculture.
- Technical specification for R&D on the development of a video analytics software and hardware complex in the robotic maintenance of industrial aquaculture, featuring a description of options for installing multimedia recording equipment on the physiological state and behavior of fish, activation tools responsible for the technological processes of fish farming operations, taking into account the integration of image processing software modules in predicting fish diseases and automating feeding.
// review — work done
The following activities and activities are planned for the first year of the project:
- Preparation of an analytical report on methods, models, algorithms, and software and hardware for video analytics in the robotic maintenance of industrial aquaculture, described in the scientific and technical literature of 2015-2025, featuring a chronological analysis of the dynamics of research on current issues and proposed approaches, studying domestic and foreign work available in publishing houses and aggregators of scientific information: RSCI, Elsevier, Springer Link, MDPI, Google Scholar, etc.
- Formation of the structure of hardware modules of an automated complex that provides registration of multimedia information about the physiological state and behavior of fish, activation tools responsible for the technological processes of fish farming operations, ensuring the integration of image processing software modules for predicting fish diseases and automation of feeding. - Development of the structure and the first version of an annotated dataset of fish images for studying the physiological state and behavior of individuals and groups of fish in cages and closed-circuit water supply installations, providing disease detection and feeding intensity management.
- Development of algorithms and neural models for preprocessing underwater images, trained on the basis of freely available data sets, characterized by a combination of image scaling and normalization methods that ensure high processing accuracy in difficult conditions with low illumination and turbidity of the aquatic environment.
- Development of algorithms and neural models for fish detection, tracking the movement of a group of fish, trained on the basis of freely available data sets, characterized by labeling objects in each frame and filtering overlapping or duplicated areas of fish detection, ensuring continuous tracking of each individual on a sequence of frames in a video recording using unique identifiers.
- Preparation of an analytical report on methods, models, algorithms, and software and hardware for video analytics in the robotic maintenance of industrial aquaculture, described in the scientific and technical literature of 2015-2025, featuring a chronological analysis of the dynamics of research on current issues and proposed approaches, studying domestic and foreign work available in publishing houses and aggregators of scientific information: RSCI, Elsevier, Springer Link, MDPI, Google Scholar, etc.
- Formation of the structure of hardware modules of an automated complex that provides registration of multimedia information about the physiological state and behavior of fish, activation tools responsible for the technological processes of fish farming operations, ensuring the integration of image processing software modules for predicting fish diseases and automation of feeding. - Development of the structure and the first version of an annotated dataset of fish images for studying the physiological state and behavior of individuals and groups of fish in cages and closed-circuit water supply installations, providing disease detection and feeding intensity management.
- Development of algorithms and neural models for preprocessing underwater images, trained on the basis of freely available data sets, characterized by a combination of image scaling and normalization methods that ensure high processing accuracy in difficult conditions with low illumination and turbidity of the aquatic environment.
- Development of algorithms and neural models for fish detection, tracking the movement of a group of fish, trained on the basis of freely available data sets, characterized by labeling objects in each frame and filtering overlapping or duplicated areas of fish detection, ensuring continuous tracking of each individual on a sequence of frames in a video recording using unique identifiers.