Prediction of the weight of mussel and pipi meat using machine vision

Institute of Food Technologists (IFT) Annual Meeting & Food Expo, Las Vegas, United States Of America, 25 - 28 June 2012, pp.252

Publication Type: Conference Paper / Full Text
City: Las Vegas
Country: United States Of America
Page Numbers: pp.252
Çanakkale Onsekiz Mart University Affiliated: No

Abstract

1. Justification

Greenshell mussels are an important exported seafood in New Zealand (more than $NZ 200

Million). One form of the product is frozen meat. Pipis are a valued bivalve in New Zealand. The estimation of the weight of the meat would allow accurate classification of these valuable

products. There are visual attributes that are important for quality for both these shellfish, either

in the shell, or shucked. Therefore machine vision (MV) can perform many operations

simultaneously for this industry.

2. Objectives

Our objective was to determine the correlation between the calibrated view area of the meat and their weight.

3. Methods

Greenshelled mussel (Perna canaliculus) and pipis (Paphies australis) were obtained from a

local market. They were alive. The shellfish were shucked, and the meat was placed in a lightbox to take pictures using a Nikon D300 digital camera. Polarized lighting was used. A reference square of known area was placed in the light box, and was present in every picture. The meat was then weighed. Correlations were developed using the data of weight vs calibrated view area.

4. Results

The linear and power curves were fitted, where X was the calibrated view area in cm2, and Y wasthe weight in g. The linear fits had an R2 > 0.92, therefore this correlation was selected since it is simpler.

5. Significance

The possibility to automate the classification of greenshell mussel and pipi meat has been

demonstrated. The MV system is a viable technique for prediction of weight of bivalves. This

technique can be used economically by the greenshell mussel and pipi industry.