Abstract:
"
Sri Lanka’s agricultural sector, based on major crops such as Tea, Rubber, and Coconut, has
always been a direct contributor to the national economy. Out of these, Sri Lankan Tea has
huge popularity and demand in the world and hence more foreign exchange is coming into the
country. Black Tea, one of the most popular teas in the world, is produced in Sri Lanka. The
cultivation in Sri Lanka, which is of such great value, is currently declining due to the
deterioration of quality due to Tea Diseases. Among them, Tea Stem & Branch diseases play a
major role. Tea Stem & Branch diseases adversely affect the general condition of the tea plant
and cause symptoms such as reduced yield and stunted plant growth. It is also possible to
contact domain experts in the field of tea cultivation to control these diseases, but it is a very
difficult task as they are very limited.
The system was created using image processing and CNN (Convolution Neural Network) to
identify tea stem and branch diseases and minimize the damage caused by them. The dataset
was created by the author under the guidance of domain experts in the Tea Research Institute.
The images selected for the dataset were adjusted correctly by changing the image backgrounds
and sizes. The dataset, which was created using pre-processed images, was mounted to a model
and then trained.
The model was trained several times and its test accuracy was tested and the model was trained
until a very good test accuracy was obtained. The trained model showed an accuracy of 94%.
It is therefore clear that the use of CNN-based algorithms for image classification is more
appropriate. Further expanding the database will make the diagnosis of tea stem and branch
diseases more effective"
