A Framework Leveraging Robotics and Machine Learning Technologies for Early Disease and Pest Detection in Greenhouse Tomato Crops

We present an integrated framework for intelligent automated plant monitoring, towards early disease and pest detection in greenhouse tomato crops. The framework combines the use of a robotic mobile platform to autonomously collect multi-spectral images of the plants, with a tool that utilizes Faster R-CNN to detect regions that signify the presence of a disease or pest. The robot is based on a modified mobile vertical mast lift platform, and integrates a 6-dof robotic arm that is used to position the plant imaging multi-spectral camera. The robot can navigate autonomously inside the greenhouse via a magnetic guidance sensor. Results from a series of experiments demonstrate the validity and effectiveness of the implemented framework.