Deep learning for crop information extraction from multi-spectral images

Subject description:

Precision agriculture is a subject of interest in the science and technology of XXI century. There are several startups developed and raising in the last years, for instance the Israeli precision agriculture startup called Taranis.

Two of the main components of precision agriculture are: a) autonomous aerial robots aka drones; and b) vision systems endowed with artificial intelligence. This thesis proposal relies on both of them: developing algorithms for extract valuable information from crop images.

This subject at CIO has started in 2017 with the Master thesis of Andrés Montes de Oca.

The presented proposal will be developed in collaboration with PhD, master and undergraduate students all members of the LAB.

Objectives:

• Developing an algorithm using CCN to identify automatically crop disease and hydric stress from multispectral and RGB images.
• Conduct experiments using the quadrotor platform already developed at the LAB.
• Publish a paper in a prestigious international conference such as: ICRA, IROS, ACC, CVPR, ICUAS.
• Attending the conference.
• A draft for an International Journal.
• A thesis

State of the project

Project started in 2017.

Actual students involved in the project:

MSC Andrés Montes de Oca

Publications from the team:

Low-cost multispectral imaging system for crop monitoring. A. Montes de Oca ; L. Arreola ; A. Flores ; J. Sanchez ; G. Flores. 2018 International Conference on Unmanned Aircraft Systems (ICUAS); pp. 443 – 451; 2018.

Materials available at the LAB:

Multispectral professional camera RedEdge; multirotor drones; JetsonTX2; professional servers for training artificial intelligence algorithms; monocular cameras.

Financial support:

• Consorcio de Inteligencia Artificial
• Consorcio para la Innovación y Transferencia Tecnológica para el Desarrollo Agroalimentario del Estado de Aguascalientes