Automated Plant Disease Diagnosis in Apple Trees Based on Supervised Machine Learning Model

The United States is the second largest producer of apples in the world with an estimated $21 billion downstream revenue. Since agriculture in the USA is highly mechanized, it is critical that latest advancements in technology are always integrated to the agricultural sector to not only improve efficiency but also improve quality, quantity, and to ensure faster distribution. Crop disease hampers the overall agricultural productivity and for a temperature-controlled crop like apple trees, identification of diseases at beginning stage is of paramount importance. There are two ways to identify and rectify issues relating to apple tree diseases, firstly by engaging expert biologists and secondly via automated identification through image processing. The biggest challenges with identification of diseases via biologist are accuracy, time constraints in case of bigger farms and budgetary limits. This research proposes the use of Machine Learning (ML) technique to aid and assist in automated disease detection and identification, and hence, making it affordable. It proposes the use of an ensemble (via weighted average) over single models, thereby improving performance and robustness by utilizing augmentations (positional and colour) which were not present in earlier studies. The proposed work surely creates an impact on the current plant disease diagnosis field by making the classification mode accurate and robust since it reaches accuracy of ~95% for all the classes.