In-season field level yield forecasting

Patent metadata

Field	Value
Publication number	US-11574465-B2
Application number	US-201916725869-A
Country	US
Kind code	B2
Filing date	Dec 23, 2019
Priority date	Dec 21, 2018
Publication date	Feb 7, 2023
Grant date	Feb 7, 2023

Abstract

In an embodiment, digital images of agricultural fields are received at an agricultural intelligence processing system. Each digital image includes a set of pixels having pixel values, and each pixel value of a pixel includes a plurality of spectral band intensity values. Each spectral band intensity value describes a spectral band intensity of one band among several bands of electromagnetic radiation. For each of the agricultural fields, spectral band intensity values of each band are preprocessed at a field level using the digital images for that agricultural field resulting in preprocessed intensity values. The preprocessed intensity values are provided as input to a machine learning model. The model generates a predicted yield value for each field. The predicted yield value is used to update field yield maps of agricultural fields for forecasting and can be displayed via a graphical user interface (GUI) of a client computing device.

First claim

What is claimed is:

1. A computer-implemented method comprising:
receiving, at an agricultural intelligence processing system, a plurality of digital images of a plurality of agricultural fields, each of the agricultural fields represented by one or more digital images obtained in-season or in different planting seasons, each digital image of an agricultural field comprising a set of pixels with pixel values, each pixel value of a pixel representing a set of spectral band intensities, each spectral band intensity of a set of spectral band intensities represented by a spectral band intensity value of one band among a plurality of bands of electromagnetic radiation;
detecting one or more cloud-covered fields among the plurality of agricultural fields, a cloud-covered field having a pixel count exceeding a threshold cloud coverage level;
removing all digital images of the detected one or more cloud-covered fields from the digital images of the plurality of agricultural fields; and then,
for each of the agricultural fields, using all remaining digital images for that agricultural field, preprocessing spectral band intensity values for each band among the plurality of bands, resulting in storing a plurality of preprocessed intensity values at a field level for all the agricultural fields;
inputting the stored plurality of preprocessed spectral band intensity values for a particular field to a trained machine learning model to result in obtaining output comprising a predicted yield value for that particular field or an agricultural field other than the particular field; and
based on the predicted yield value, causing generating and displaying an updated field yield map of the particular field using a graphical user interface of a client computing device.

2. The method of claim 1 , wherein the preprocessing further comprises:
calculating an aggregated mean or histogram value of the spectral band intensity values of all digital images of the particular field for each band.

3. The method of claim 1 , further comprising:
applying the predicted yield value to one or more digital images of the particular field or the agricultural field other than the particular field in a subsequent planting season.

4. The method of claim 1 , further comprising:
re-training the trained machine learning model with digital images of another field obtained during a subsequent planting season, the predicted yield value, or a combination thereof; and
periodically repeating re-training the trained machine learning model with a set of digital images obtained of the particular field, one or more agricultural fields other than the particular field, or a combination thereof, during subsequent planting seasons.

5. The method of claim 1 , further comprising:
grouping the predicted yield value with a particular batch of a plurality of batches of sets of predicted yield data corresponding to a set of fields other than the particular field, each batch with a set of predicted yield data;
inputting the particular batch to the trained machine learning model; and
repeating the inputting function for batches other than the particular batch.

6. The method of claim 1 , wherein predicted yield data of batches of sets of predicted yield data correspond to one or more regions of interest within an agricultural field.

7. The method of claim 1 , wherein the plurality of bands of electromagnetic radiation includes at least Red, Green, Near-Infrared (NIR) and Red-Edge (RE).

8. The method of claim 1 , further comprising applying one or more of image band normalization, temporal averaging, or determining location cross features to prepare features provided to the trained machine learning model.

9. A computer-implemented method comprising:
receiving, at an agricultural intelligence processing system, a plurality of digital images of a plurality of agricultural fields, each of the agricultural fields represented by one or more digital images obtained in-season or in different planting seasons, each digital image of an agricultural field comprising a set of pixels with pixel values, each pixel value of a pixel representing a set of spectral band intensities, each spectral band intensity of a set of spectral band intensities represented by a spectral band intensity value of one band among a plurality of bands of electromagnetic radiation;
for one or more agricultural fields with missing digital images, imputing a spectral band intensity value, per band, for each of the spectral band intensity values of a pixel of the set of pixels of digital images;
flagging the one or more agricultural fields with missing digital images with a missing image indicator(s);
for each of the agricultural fields, using all the digital images for that agricultural field, preprocessing spectral band intensity values for each band among the plurality of bands, resulting in storing a plurality of preprocessed intensity values at a field level for all the agricultural fields;
inputting the stored plurality of preprocessed spectral band intensity values for a particular field and the missing image indicator(s) to a trained machine learning model to result in obtaining output comprising a predicted yield value for that particular field or an agricultural field other than the particular field; and
based on the predicted yield value, causing generating and displaying an updated field yield map of the particular field using a graphical user interface of a client computing device.

10. The method of claim 9 , wherein the preprocessing further comprises:
calculating an aggregated mean or histogram value of the spectral band intensity values of all digital images of the particular field for each band.

11. The method of claim 9 , further comprising:
applying the predicted yield value to one or more digital images of the particular field or the agricultural field other than the particular field in a subsequent planting season.

12. The method of claim 9 , further comprising:
re-training the trained machine learning model with digital images of another field obtained during a subsequent planting season, the predicted yield value, or a combination thereof; and
periodically repeating re-training the trained machine learning model with a set of digital images obtained of the particular field, one or more agricultural fields other than the particular field, or a combination thereof, during subsequent planting seasons.

13. The method of claim 9 , further comprising:
grouping the predicted yield value with a particular batch of a plurality of batches of sets of predicted yield data corresponding to a set of fields other than the particular field, each batch with a set of predicted yield data;
inputting the particular batch to the trained machine learning model; and
repeating the inputting function for batches other than the particular batch.

14. The method of claim 9 , wherein predicted yield data of batches of sets of predicted yield data correspond to one or more regions of interest within an agricultural field.

15. The method of claim 9 , wherein the plurality of bands of electromagnetic radiation includes at least Red, Green, Near-Infrared (NIR) and Red-Edge (RE).

16. The method of claim 9 , further comprising applying one or more of image band normalization, temporal averaging, or determining location cross features to prepare features provided to the trained machine learning model.

17. A system comprising:
one or more processors;
a memory storing instructions which, when executed by the one or more processors, cause performance of functions comp

Assignees

• Climate Llc

Inventors

• Chen Yaqi
• Johannesson Gardar
• Guan Wei

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06V10/143

  Sensing or illuminating at different wavelengths · CPC title

• G06V20/194

  using hyperspectral data, i.e. more or other wavelengths than RGB · CPC title