Line Laser Scanning Combined with Machine Learning for Fish Head Cutting Position Identification.

Area of Science:

• Agricultural Engineering
• Computer Science
• Data Science

Background:

• Manual fish head cutting relies on experience, hindering large-scale production.
• Automated solutions are needed to meet industrial demands for efficiency and accuracy.

Purpose of the Study:

• To develop a fast, contactless method for identifying fish head cutting positions.
• To establish and compare machine learning models for accurate cutting point identification.

Main Methods:

• A line laser data acquisition system was constructed to collect fish surface data.
• Principal Component Analysis (PCA) reduced data dimensions.
• Long and Short Term Memory (LSTM), Least Squares Support Vector Machines (LS-SVMs), and Particle Swarm Optimization-Back Propagation (PSO-BP) networks were applied for model development.

Main Results:

• The LSTM model achieved the highest accuracy with R2=0.9480, RMSE=0.2957, MAE=0.1933, and RPD=3.1426.
• The developed system demonstrated reliable and rapid identification of fish head cutting positions.
• Machine learning integration with laser scanning proved effective for automated fish processing.

Conclusions:

• The LSTM model offers a robust solution for automated fish head cutting position identification.
• This technology can guide the development of intelligent fish processing and cutting equipment.
• Combining line laser scanning with LSTM enhances precision and speed in fish pre-processing.