Automated approach for oil palm in vitro shoot classification

Oil palm tissue culture shows promising future in providing uniform and quality cloned ramets for planting. However, mass production of oil palm plantlet is currently prohibitive in spite of its high demand. This is due to the fact that most of the processes in tissue culture are done manually which is labour-intensive as well as prone to contamination. In order to mass-produce clonal planting materials, an automated system is desirable, which is expected to be more cost effective as well as enhancing efficiency. In this study, the automation system is targeted at the shoots development stage since it is tedious job and routine to be operated manually as no automation system is currently adaptable for the tasks. This research focuses on the classification of various categories of normal and abnormal oil palm in vitro shoots. The oil palm in vitro shoots samples were digitized and employed for automated approach. A customized method for automatic image thresholding has been proposed to deal with various samples geometrical orientations and a wide variety of lighting conditions, as the environment for the future automation system would likely to be this way. Features were later extracted based on thinning and convexity image morphologies. By manipulating the features data obtained, three classification methods have been experimented, namely Linear Discriminant Analysis, K-mean clustering and back-propagation neural network. Results showed that all classification methods perform well, and able to differentiate between normal and abnormal oil palm in vitro shoots, with highest classification rate at ninety three percent. This is expected to greatly facilitate the development of the prospective automation system.

Preview PDF FK 2010 10R.pdf Download (737kB) | Preview

Actions (login required)

View Item