There is currently a significant trend towards integrating automation with sensory data to develop precision agriculture systems. These systems can lead to increased productivity at a lower cost and a reduction in waste. This thesis presents an integrated architecture of a data-driven robotic system for plant monitoring in tomato greenhouses. Greenhouses represent an excellent case study for the impact of automation due to the high cost of labor and vegetation intensity. The proposed architecture focuses on monitoring at the individual plant level, enabling significant potential improvements in disease scouting, yield prediction, and automated labor supervision. The design is comprised of a mobile robotic data collection platform, for consistent and dense data collection, and hierarchical deep learning architecture for automated analysis of plant images. The robotic platform was evaluated using teleoperated and semi-autonomous modes, and the design included features for full autonomy. The hierarchical scene understanding system incorporated capabilities for the detection and segmentation of plant morphology, including leaves, stems, and tomatoes, disease detection, and plant tracking. Disease detection and severity estimation of individual leaves showed that symptoms are largely transferable between pathogens, enabling the detection of previously unseen conditions. Cluster tracking was accomplished by employing a detect-to-track technique using the tomato detection system and a general object tracker. Annotation tools were developed for the efficient labeling of data for use in machine learning training procedures. Additional work is required to fully automate the system, and for practical deployment in commercial greenhouses.