Method of visual detection of plant diseases in landless indoor agriculture environments (MESOI)

Context:
Research project by the Research Group on Media Technologies at La Salle – Universitat Ramon Llull, where I worked as a research assistant, contributing to system development, result analysis, and the writing of reports and publications.
Technologies:
Computer Vision · Deep Learning · Python · Matlab · Mobile Robotics (TurtleBot, Jetson TX1)
Description:
This project focused on developing an automated monitoring system for indoor agriculture using a mobile robot equipped with a camera and computer vision capabilities. The system aimed to detect plant diseases and monitor crop growth autonomously, reducing the need for manual inspections.
We captured image datasets from local agricultural fields—labeling images of leaves, pests, diseases, and fruits—and combined them with publicly available datasets. These were used to train deep learning models for object detection and classification, evaluating their accuracy in real agricultural scenarios.
The system was ultimately tested on a TurtleBot mobile robot equipped with a Jetson TX1 board and camera, navigating a simulated field to perform detection tasks in real time.
As part of my work, I:
- Captured and labeled large image datasets for training and evaluation
- Designed, trained, and tested deep learning models for image classification
- Programmed a mobile robot to autonomously collect and analyze visual data
- Collaborated with biologists and farmers to understand plant health indicators and validate the system’s practical use
- Authored technical reports and documentation, including dataset records and experiment results
Project Website salleurl.edu/en/research/project/mesoi
Video of the simulated field experience: