World Allergy Congress (WAC) LYON, France, 12-14 décembre 2019
acquisitions images are realized in an infested maize field. They, in situ, allow analyses of
sensitivity of spatial resolution and spectral characteristics according to the crop, to the
invasive plant infestation levels. This experiment focuses on spatial resolution studies made
possible by the Very High Resolution drone images that are compared to sentinel-2 images.
The aim in the future would be a complementary drone/satellite evaluation for modeling
The study area is located in the suburb of LYON: Saint Priest, France. The drone used is the
Parrot senseFly ebee equipped with the Sequoia sensor. The flight took place on the
2017/05/29 (10h31). The spatial resolution of images, acquired in four bands (green, red, red
edge and NIR) is 6 cm pixel size (ps). The drone image is compared to sentinel-2 images (ps:
10 m, 20 m) acquired on cloud-free 2017/06/02. Field surveys define ROIs identifying 9 land
use classes, 1: ragweed, 2: goosefoot, 3: bare soil1, 4: bare soil2, 5: tree, 6: tree shade, 7:
maize, 8: maize inter-rank. The bare soils 1 and 2 are differentiated by their color.
The ragweed reflectances measured by the drone are different in other land uses. They are
weak in the red, highest in the red edge, high and identical to the reflectances of trees in the
NIR. The results of the classification carried out on all the channels and the NDVI are good.
The percentage of correctly classified pixels is 84%, if the two bare soils classes are grouped.
To test the influence of the sentinel-2 spatial resolution, images are re-sampled by the nearest
neighbor method at lower resolutions. The quality of the classification is maintained until the
resolution 1 m/ps.
This acquisition by drone enables ragweed sensitivity studies at an early stage of growth
period of maize plants. Under the conditions of the study (maize height 80 cm and 30-40
plants in 1 m2, number of ragweed plants between someone and more than 20/m2), the spatial
resolution must be at least 1 m.