Classification of Phalaenopsis Plantlet Parts and Identification of Suitable Grasping Point for Automatic Transplanting Using Machine Vision

Abstract

This study develops an image-processing algorithm to segment and classify the components of Phalaenopsis tissue culture plantlets (PTCPs) and to determine a suitable grasping location on the roots for an automatic transplanting operation. The algorithm uses the nodes of the plantlet's skeleton to generate cutting lines to separate the plantlet into its constituent leaves and roots. A Bayes classifier based on an optimal combination of color and shape features is then applied to classify the individual segments of the plantlet as either leaf or root segments. The root segment with the highest decision value based on the Bayes theorem is then selected, and the midpoint of its skeleton specified as a suitable grasping point for an automatic transplanting operation. The classification results obtained by the Bayes classifier for manually cut samples demonstrate that a classification accuracy of 99.9% is achievable given an appropriate choice of color and shape features. Implementing the optimal set of features, the proposed classifier achieves a 94.9% success rate in identifying suitable grasping points on complete PTCP plantlets. Therefore, the experimental results indicate that the proposed image-processing algorithm has the potential for integration with a robotic handling device to realize an automatic machine vision plantlet transplanting system.