Fish Identification via Coupled Texture and Geometric Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for lossless biomass measurement of industrially farmed fish are inefficient, as they rely on fuzzy measurements that can lead to overfitting or local optimization, and existing fish face identification techniques using convolutional neural networks have low accuracy and require a large number of samples.
Innovation Solution
A method and system for fish identification based on body surface texture features and geometric features, using an improved ResNet network and a deep learning YoLo network to extract and couple these features, allowing for identity recognition of individual fish with small sample learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a simple convolutional neural network is used for fish face identification, then the training process is simpler, but the identification accuracy is low and a large number of samples are required
Solution Approach 1:
The patent combines multiple feature extraction methods (body surface texture features using improved ResNet and geometric features using deep learning YoLo network) into a unified identification system. This merging of multiple feature types enables high-accuracy identification while maintaining the ability to train with small samples, thus resolving the contradiction between training simplicity and identification accuracy.
2Ease of operation
If traditional fuzzy measurement methods are used for biomass measurement, then the measurement process is simpler, but the measurement accuracy is low and overfitting occurs
Solution Approach 1:
The patent changes the measurement parameters from traditional fuzzy measurement to precise individual identification based on body surface texture and geometric features. By extracting and coupling these specific parameters, the system achieves lossless accurate biomass measurement while avoiding overfitting through small sample learning capabilities.
3Quantity of substance
If repeated measurement of the same fish is performed, then more data is collected, but overfitting and local optimization occur
Solution Approach 1:
The patent implements identity recognition as a feedback mechanism to identify and eliminate duplicate measurements of the same fish. By recognizing individual fish identities through coupled feature analysis, the system prevents repeated measurement of the same object, thereby avoiding overfitting and local optimization while maintaining measurement model reliability.
Data Source
AI summary
A method and system for fish identification based on body surface texture features and geometric features are provided. The method employs an improved Resnet network and a deep learning Yolov8 network to extract body surface texture features and geometric features of a fish on the basis of considering influences of fish tail swing and an oxygen concentration on a fish body form based on a small sample learning framework, and then realizes identity recognition of a fish individual by coupled analysis of the body surface texture features and the geometric features. The method can realize high-accuracy fish identification with few training samples of a fish to be identified from the perspective of actual application, provides theoretical basis and technical support for accurate fish stock assessment and accurate estimation of industrially farmed fish biomass, and meets the development requirements of modern agriculture.