Pink and chum salmon. Photo credit: Scott Smiley.
Machine Vision System for Seafood Quality Control
Problem: One of the hurdles facing many Alaska seafood processors is the consistency of quality expected by many buyers. In today’s highly competitive food markets, buyers insist on an unvarying level of quality. Commercial users of Alaska seafood will project yields and recoveries, price points, and profitability based on product specifications. When these are not met, the buyers are faced with loss of product, market and money. Meeting these specifications often require careful evaluation and grading of raw and finished material. This is achieved using trained graders and quality control technicians who make subjective judgments on defect levels and quality criteria. Unfortunately, these judgments can vary from grader to grader, from day to day and from lot to lot, making consistent quality hard to achieve. This project aimed to develop more objective methods to improve the consistency of Alaska seafood quality.
Solution & Approach: This project is ongoing. We built a vision system using a video camera, lighting system, digital capture card and software developed at the University of Florida. Once the vision system was assembled, we determined optimum-operating conditions. Finally, product properties were investigated and the vision system was used to evaluate its potential as a quality control tool. The final task was to test the vision system in a processing plant. We tested the system as a quality control tool for visualizing defects in fillet operations, sorting salmon by meat and skin color, identifying species of rockfish and flatfish and evaluating the quality of incoming raw material. Products were run through the vision system as well as the traditional methods to compare the differences in quality.
Principal Investigator: Alexandra de Oliveira