Detection of White Stem Borer Disease in Coffee Plantation using Autonomous Multi Terrain Robot

Abstract

A number of nations’ economies rely heavily on the production of coffee. The coffee plant has experienced significant damage as a result of a White Stem Borer (WSB) bug. The WSB bores into the coffee stem and stays buried there until the plant fully falls. As it is challenging to diagnose the disease in the early stages of its infestation, this disease greatly distresses coffee growers. In order to find a way to prevent the significant yield loss, we built on a dataset of WSB-infected stems, identified 3 distinct features to classify infected coffee stems and distinguish them from healthy stems, and then trained YoloV5 deep learning model. We developed a model using augmented images from the existing image data after dividing it into training and testing sets. mAP, precision, and recall of 81.9, 89.7 and 73.8 percentages respectively were obtained, which would increase substantially with larger data size. The proposed method of utilising the bark’s characteristics for early detection of stem borer disease has not been explored so far. The image processing pipeline implemented is accompanied by the simulation of Arabica Coffee plantationlike environments. The simulations focus on the Jackal UGV as it is an ideal robot for the use case of autonomous deployment in mapping and navigation of the plantations. Functionalities such as Elevation Mapping, State Estimation and Localisation have been incorporated in tandem so that data captured by use of a depth camera can be used efficiently for operation.