Speaker
Description
The use of a double-row pineapple sucker planter increases the efficiency of pineapple sucker planting operations compared to conventional methods that labour intensive and high time consuming. However, there are several suckers planted on the soil that do not meet the characteristics of good planting quality for optimal pineapple crop growth when the planting machine is used. After every planting session, farm workers are required to inspect the position of the planted sucker. This causes the use of the planting machines to be less efficient because additional workers and time are needed to inspect the suckers planting position. For that reason, a model was developed to classify the sucker planting quality in real-time using the Convolution Neural Network (CNN) technique. The developed model classifies the sucker planting quality into two classes, namely the OK class and the misplant class. In this study, 2000 sucker image samples were captured for data acquisition. The annotated and labeled image samples were used to train the sucker planting quality classification model. Based on the confusion matrix obtained, the accuracy of the developed model was 86%. The model was then deployed at the planter machine and classification of the sucker planting quality was performed in real-time while the planting operation was carried out.
