Date of Submission

6-2021

Document Type

Thesis

Degree Name

Master of Science in Mechanical Engineering

Department

Mechanical and Industrial Engineering

Advisor

Cheryl Li, Ph.D.

Committee Member

Maria-Isabel Carnasciali, Ph.D.

Committee Member

Eric Dieckman, Ph.D.

Keywords

Fish Processing, Fishing Industry, Artificial Intelligence, Machine Learning

LCSH

Fisheries, Fishery processing, Fishery processing plants, Artificial intelligence--Agricultural applications, Machine learning

Abstract

This thesis presents an approach to automatic fish processing in the fishing industry using artificial intelligence. Using transfer learning approach (VGG16 model), convolutional neural network (CNN), and image processing, this thesis demonstrates the series of steps from intake of fishes, chopping of head and tail to collecting the body parts of fishes separately thus making it ready for post-processing or direct packaging as per the industrial and consumer needs. To present the idea of the classification of fishes, we have considered five different fishes and one shrimp, altogether six classes. After the classification, the fishes are sorted out according to their classes and loaded on their respective conveyors through graders. Fishes that are smaller than threshold thickness are sorted and separated through the grader for manual work. The remaining fish proceeds in the conveyor line through an inclined pocket conveyor and each pocket is designed to hold one fish. This is done to achieve separate the fish individually to ease the post-processing operations. The length and width are determined through conventional image moment algorithm and center of mass through a deep neural network. Again, the fishes weighing and measure above the threshold weight and length are removed automatically with guide flaps for manual work. The orientation of the fish is then adjusted to a predetermined position (head always in one direction) by a robotic hand (R1) using the center of mass co-ordinates followed by chopping operation. This operation is performed after the fish is brought to a correct position and orientation followed by determination of head and tail chopping points. These two operations before cutting are accomplished through two deep neural networks trained to determine the correct position and chopping locations, respectively. The chopping assembly adjusts itself with dimension, the center of mass, and chopping points of fish before loading of fish in the cutter. The second robotic hand (R2) picks the fish, place it in cutting assembly, and blades actuates separating and collecting the head, tail, and body in different containers. The training and test accuracy of all the networks are promising and above 90%. Reducing the need of human involvement around heavy machines, this approach increases the human safety factor. Using artificial intelligence in the whole fish processing also makes the operation safe with the least human involvement, more robust, and more economical. Handling the major processing works like sorting, grading, staging, and preliminary cutting ease the remaining complex procedures which will be handled by the human workforce thus making the whole process faster and productive.

Download

Share

COinS