The development of an automatic rubber seed sowing system with machine vision assistance

• Published in: Journal of rubber research Vol. 25; no. 3; pp. 187 - 194
• Main Authors: Mustafah, A. Mohd, Khairunniza-Bejo, S., Lim, Y.
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.08.2022
• Subjects: Ceramics; Chemistry and Materials Science; Composites; Glass; Materials Engineering; Materials Science; Natural Materials; Original Paper; Plant Biochemistry; Polymer Sciences; Robotic arm; Machine vision; Rubber sowing; Rubber seed
• ISSN: 1511-1768; 2524-3993
• DOI: 10.1007/s42464-022-00153-8

Natural rubber is an important world commodity to produce gloves, tyres and many other important goods. Rubber plant naturally grows from seeds and it can also be grafted artificially. Rubber seedlings in nurseries are currently manually sowed by labour. The rubber seed must be placed with the dorsal side facing upwards and the ventral side facing downwards. Due to the labour shortage in plantations currently faced in Malaysia, nursery automation would be beneficial to maintain production and to increase productivity. An automated nursery seeding system would be able to reduce labour requirements and further increase the seed placement accuracy of manual sowing. The objective of this paper is to elaborate on the development and testing of an automated robotic arm with camera vision assistance for sowing rubber seeds. The developed robotic arm was shown to successfully pick rubber seeds and place them in a polybag germinator in the correct orientation. An integrated machine vision was used to assist in identifying the surface of rubber seeds to get the correct positioning during sowing. The seed image frames were obtained from a recorded video from a smartphone camera and they were transferred to a computer via a wireless network and processed using Matlab to identify the dorsal and ventral side of the seed. The developed system had managed to achieve 90% accuracy under indoor light condition of 40 samples of rubber seeds tested. The system gave the highest precision (100%) at 50% motor speed setting. The average time of 10 s was taken to complete the operation from capturing the seed image to successfully placing the seed in the correct orientation.