Underwater Camera Biomass Forecasting for Non-Invasive Fish Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The manual process of removing and weighing fish from fish pens in aquaculture is time-intensive and potentially harmful, and it only provides limited insight into the true characteristics of the fish population, making it difficult to manage feeding and harvesting effectively.
Innovation Solution
An underwater camera system coupled with computer vision and machine learning techniques is used to capture images of fish, process them to identify individual fish, and predict biomass distribution using models like neural networks, allowing for automated feed control and decision-making based on projected biomass changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual fish removal and weighing is performed, then biomass data can be obtained, but the process is time-intensive and harmful to fish
Solution Approach 1:
The patent replaces the mechanical manual process of removing and weighing fish with an optical imaging system. Underwater cameras capture images of fish in their natural environment, and computer vision algorithms automatically measure fish dimensions and estimate biomass, eliminating the need for physical fish handling and manual weighing operations.
Solution Approach 2:
The patent introduces an intermediary computational system between the fish population and the biomass measurement. Machine learning models and computer vision algorithms serve as intermediaries that process visual data to infer biomass characteristics without direct physical interaction with the fish, thereby avoiding harm while obtaining accurate measurements.
2Measurement precision
If manual fish sampling is performed, then some biomass data can be collected, but only a small portion of the population is measured
Solution Approach 1:
The imaging system is designed to simultaneously capture and analyze multiple fish within a single frame or sequence of images. The system can process entire school populations or large subsets of the population, providing comprehensive biomass data that represents the whole population rather than just a small sampled portion.
Solution Approach 2:
The patent creates visual copies (images) of the fish population that can be analyzed without altering or removing the original fish. These digital copies allow for repeated measurement and analysis of the same population multiple times, providing comprehensive data coverage of the entire population.
3Object-affected harmful factors
If automated image processing is implemented, then fish population monitoring becomes non-invasive, but system complexity increases
Solution Approach 1:
The patent extracts the measurement function from the physical fish handling process and relocates it to a computational analysis system. By separating the observation function (imaging) from the measurement function (computer vision analysis), the system eliminates harmful physical contact while concentrating the complexity in software rather than mechanical systems.
4Measurement precision
If computer vision and machine learning models are used, then biomass prediction accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The patent implements preliminary action by pre-training machine learning models with extensive fish data before deployment. The models are pre-trained to recognize fish species, estimate dimensions, and predict biomass from visual features. This preliminary training enables rapid inference during actual monitoring operations, reducing real-time processing time while maintaining high accuracy.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for underwater camera biomass prediction. In some implementations, an exemplary method includes obtaining one or more images of a population of fish captured by an underwater camera; providing data corresponding to the one or more images to a model trained to predict biomass values; obtaining output of the trained model including a predicted biomass value indicating a future biomass of a fish within the population of fish; and determining an action based on the predicted biomass value.