Remote sensing by embedded vision on an autonomous UAV for precision agriculture

Abstract

Farming is an essential work for the humankind and the main source of food. Precision farming is the beginning of a bigger project which is the farm of the future. It will use fewer labor workers better energy and water consumption resulting in high production, lower time and cost. In our topic we will focus on using drones to analyze agricultural information where it will be used to make decisions, for example controlling the water supply if the plants are weak, fruit counting, launching the piking process using an unmanned ground vehicle UGV if the products are ready . The mapping of an agricultural field based on the NDVI vegetation index is not always sufficient to identify areas of contamination or diseases detected in the vegetation. It sometimes requires the collection of samples in the field, for a more precise analysis in ex-situ. This project consists of equipping a quad-rotor drone made in a former PFE, an Odroid XU-4 electronic card, a camera, and an GPS to perform the task of taking samples in the field. Visual feedback and GPS data can position the platform on visual targets with precision (± 1cm). The optimal trajectory which passes through all the points of interest (way-points) avoiding obstacles is established by the application of CPP (Coverage path planning) where the vehicle must intelligently plan it trajectories while avoiding existing obstacles in the environment then proceed to fruit counting to estimate the necessary storage. Google maps is great for general view of an area but the lack of resolution and update rate makes it unusable for precision agriculture, in this project we are able to use alternative maps with higher resolution and even apply coverage path planning algorithms on them and also create a flower counting mission at the end