Design of a Real-Time Silkworm Cocoon Sorter Using Machine Learning

Silk is known for its luster, softness and strength, making it a prized material for textile and other consumer applications. Silk is a natural protein fiber produced by silkworm cocoons and the quality of the raw silk produced directly depends on the quality of the cocoons used for reeling. Higher quality cocoons fetch a better price in the cocoon market and sorting of cocoons into good and bad becomes crucial to produce high quality silk. As of now, cocoon sorting is carried out manually and with a drastic increase in the number of cocoons produced and received at silk reeling centers, there is a need to automate cocoon sorting to assist humans. In this paper, machine learning based real-time cocoon sorting system using multiple sensors is proposed. Two machine learning algorithms, decision tree model and Neural Network model running on an ESP32 module and ESP32 camera module respectively are used for the proposed sorting machine to sort good quality cocoons from low quality cocoons. The proposed system is designed at a cost of $ 250 and is capable of sorting cocoons ranging from 600 ms to 7 s depending on cocoon type with a maximum recognition accuracy of 95.17 %.