公开(公告)号：EP4250250A1

公开(公告)日：2023-09-27

申请号：EP22164121.0

申请日：2022-03-24

Applicant: BASF SE

Inventor： PRIESE, Benjamin ,  KARAMI, Mojtaba ,  GEDMINAS, Laurynas ,  HAERING, Tim ,  DOERR, Manuel Josua ,  BOEHM, Julian ,  PFRANG, David ,  CHRIST, Thomas

IPC: G06V20/10

Abstract: The present invention generally relates to digital farming. In order to facilitate the provision of sensor data with a high spatial coverage, a computer-implemented method (100) is provided for training a data-driven model for measuring a concentration of a physicochemical parameter from a target field, the method comprising:
a) providing (110) historical satellite imagery data collected from at least one field over a period of time;
b) providing (120) historical sensor data collected from the at least one field, wherein the historical sensor data provides a measurement or an estimation of the concentration of the physicochemical parameter over the period of time; and
c) training (130) the data-driven model with a training dataset based on the provided historical satellite imagery data and the historical sensor data, wherein the trained data-driven model is usable as a soft-sensor for measuring a concentration of the physicochemical parameter from the target field based on satellite imagery data acquired from the target field.