Classification of Coir Fibres using Machine Vision

Abstract

Measuring the length and the width of a sample of coir fibres manually is a time consuming and labour intensive task. With the aim of developing an imaging system based on machine vision, automatic computing of length and width were investigated by processing a set of scanned images of coir fibre samples. Two algorithms were developed using a line searching technique based on the pixel arrangement, in order to measure the length and width of the fibres. Results showed that the automatically calculated lengths and actual length are linearly correlated with a correlation coefficient of 0.99. The overall error of the length calculation is ±6.4 mm. Over 75% of fibre lengths can be measured within a percentage error of 1.5%. However, the calculation of the width of the fibres did not provide satisfactory results since the estimated error value ±0.05 mm is quite high compared to the width of the fibres (∼0.3 mm). Although the percentage error is high for the estimation of the width, the results indicate that the computational technique could be improved and used as an alternative to the manual techniques.