A b s t r a c t. In the present study a machine vision system was developed for estimating the mass of eggs arranged in a single array. A grabber frame equipped with a mirror was developed for positioning the eggs. Therefore, two images could be captured from each egg. Images were then processed by Matlab software. Six algorithms were developed to extract eggs features such as minimum, maximum and effective radii, perimeter and the frontal area from each image. The eggs were also weighed by a sensitive digital scale. Seventy percent of data after discarding the outliers were used to establish some models, and the remaining was used to verify the final model. The results showed that egg mass estimation can be accurate by using two perpendicular views of each egg. Amongst the models, one with predictors of area and effective radius was found to be the best. A high correlation coefficient was observed between eggs mass measured and predicted by the model, with an accuracy of about 95%.K e y w o r d s: egg, fresh mass, machine vision INTRODUCTION Today, egg is extremely distributed in international trade, and the egg industry is a vital portion of the world food industry. In the egg trade, this product is sold by its mass. Also, many investigations have shown that egg mass can be considered as an important parameter for prediction of features of egg shell, hatchability, and chick mass (Narushin et al., 2002). Poultry products, just as other food products, have several unique characteristics which set them apart from engineering materials. These properties determine the quality of the products, and identification of correlation among those properties makes quality control easier (Jannatizadeh et al., 2008). Egg mass measurement is an essential unit operation for controlling the egg production process in the poultry industry. Information regarding egg mass is not only vital for grading systems based merely on mass, but it is also necessary for assessing quality indices such as yolk-albumen ratio, shell thickness and hatchability.Physically, weighing the individual items is very expensive and impractical. To overcome this problem, correlated mass with size is often used as a substitution for weighing each item of produce. For this purpose, machine vision is a desirable implement and can been used for size estimation. Machine vision is a non-destructive method that involves image analyses and image processing operations. Many researchers have used this method for size grading of agricultural products. Brosnan and Sun (2002), Esehaghbeygi et al. (2010), Guyer and Yang (2000), Khojastehnazhand et al. (2009) and have all found the image processing approach fast and reliable for automatic fruit sorting, defect detection and product grading.Wang and Nguang (2007) designed a low-cost sensor for measuring the volume and surface area of agricultural products. They considered each product as a set of elementary cylindrical objects of unit pixel height, and estimated the volume by summing the elementary volumes of each cylinder. ...
