Silkworm Cocoon Assessment Using GRNN on X-Rays

The source of silk is filaments from the cocoons of domesticated silkworms. The cocoons are sold in market based on weight and quality of raw silk produced from collected samples. Currently buying of silkworm cocoons depends on visual inspection carried by a human expert which does not guarantee the accuracy in grading of silk nor in classification of cocoons. This method can create favoritism towards buyers. Another method is destructive testing by using cocoon cutting and measuring raw silk produced by cocoon. This method is laborious and expensive. Hence, there is a need for a method which will have minimal human intervention and also be a non-destructive assessing method. In this paper a non-destructive testing method using X-ray technique is proposed to estimate the raw silk content, gender discrimination and classification of cocoon as normal or double using image processing technique. Extracted features are given as input to a Generalized Regression Neural Network (GRNN) to predict the weight of raw silk. The average accuracy of proposed model is 94.56% with MSE of 0.0044.. This model provides an automatic fair method for buyers and sellers to estimate the raw silk quantity and grade the silkworm cocoons precisely.