Mapping soil properties with satellite data using machine learning approaches

Patent metadata

Field	Value
Publication number	US-11468669-B2
Application number	US-201916707168-A
Country	US
Kind code	B2
Filing date	Dec 9, 2019
Priority date	Dec 11, 2018
Publication date	Oct 11, 2022
Grant date	Oct 11, 2022

Abstract

In an embodiment, a computer-implemented method for predicting subfield soil properties for an agricultural field comprises: receiving satellite remote sensing data that includes a plurality of images capturing imagery of an agricultural field in a plurality of optical domains; receiving a plurality of environmental characteristics for the agricultural field; generating a plurality of preprocessed images based on the plurality of satellite remote sensing data and the plurality of environmental characteristics; identifying, based on the plurality preprocessed images, a plurality of features of the agricultural field; generating a subfield soil property prediction for the agricultural field by executing one or more machine learning models on the plurality of features; transmitting the subfield soil property prediction to an agricultural computer system.

First claim

What is claimed is:

1. A computer-implemented method for predicting subfield soil properties for an agricultural field, the method comprising:
receiving satellite remote sensing data that includes a plurality of images capturing imagery of an agricultural field in a plurality of optical domains;
receiving a plurality of environmental characteristics for the agricultural field;
generating a plurality of preprocessed images based on the plurality of images and the plurality of environmental characteristics;
identifying a plurality of features of the agricultural field based on the plurality preprocessed images;
generating a subfield soil property prediction for the agricultural field by executing one or more machine learning models on the plurality of features;
transmitting the subfield soil property prediction to an agricultural computer system.

2. The method of claim 1 , wherein the generating of the plurality of preprocessed images based on the plurality of images and the plurality of environmental characteristics comprises: correcting atmospheric artifacts in the plurality of images; selecting, from the plurality of images, the plurality of preprocessed images that include pixels depicting snow-free bare soil.

3. The method of claim 1 , wherein the subfield soil property prediction includes information about predicted subfield soil properties of the agricultural field; wherein the information about the predicted subfield soil properties includes information about one or more of: soil organic matter content, caution exchange capacity, potential of hydrogen, potassium, phosphorous, magnesium, or calcium.

4. The method of claim 3 , further comprising: generating a plurality of subfield soil properties maps based on the information about the predicted subfield soil properties of the agricultural field; wherein the plurality of subfield soil properties maps includes one or more of: maps having different resolutions; maps depicting different soil properties, maps depicting different geographical regions, or maps depicting a geographical region at different points in time.

5. The method of claim 1 , wherein the one or more machine learning models include one or more of: a Gaussian process regression (GPR) model, a random forest (RF) model, or any other non-parametric machine learning model;
wherein the one or more machine learning models have evolved from each other and include successive versions of an original machine learning model.

6. The method of claim 1 , further comprising: identifying one or more management zones within the agricultural field based on the subfield soil property prediction for the agricultural field.

7. The method of claim 1 , wherein the plurality of images of the satellite remote sensing data includes one or more of: an image in a visible optical domain, an image in a near-infrared optical domain, or an image in a short-wave infrared optical domain.

8. The method of claim 1 wherein the plurality of environmental characteristics for the agricultural field includes one or more of: topographical data for the agricultural field, elevation, slope, aspect, curvature, and accumulated flow data, precipitation and temperature data accumulated during a time period during which the plurality of images was captured, long-term average daily precipitation and temperature values or absolute point-based soil sampling measurements of soil of the agricultural field.

9. The method of claim 1 , further comprising: generating a plurality of uncertainty and importance indicators for inputs used to generate the subfield soil property prediction for the agricultural field by executing the one or more machine learning models on the plurality of features.

10. The method of claim 1 , wherein the generating of the plurality of preprocessed images based on the plurality of images and the plurality of environmental characteristics comprises classifying images of the plurality of preprocessed images into one or more of: a no data class, a saturated pixels class, a dark features/shadows class, a cloud shadows class, a vegetation class, a bare soil class, a water class, a cloud low probability class, a cloud medium probability class, a cloud high probability class, a cirrus cloud class, or a snow/ice class.

11. One or more non-transitory computer-readable media storing one or more computer instructions which, when executed by one or more computer processors, cause the one or more computer processors to perform:
receiving satellite remote sensing data that includes a plurality of images capturing imagery of an agricultural field in a plurality of optical domains;
receiving a plurality of environmental characteristics for the agricultural field;
generating a plurality of preprocessed images based on the plurality of images and the plurality of environmental characteristics;
identifying a plurality of features of the agricultural field based on the plurality preprocessed images;
generating a subfield soil property prediction for the agricultural field by executing one or more machine learning models on the plurality of features;
transmitting the subfield soil property prediction to an agricultural computer system.

12. The one or more non-transitory computer-readable media of claim 11 , wherein the generating of the plurality of preprocessed images based on the plurality of images and the plurality of environmental characteristics comprises: correcting atmospheric artifacts in the plurality of images; selecting, from the plurality of images, the plurality of preprocessed images that include field depicting snow-free soil.

13. The one or more non-transitory computer-readable media of claim 11 , wherein the subfield soil property prediction includes information about predicted subfield soil properties of the agricultural field; wherein the information about the predicted subfield soil properties includes information about one or more of: soil organic matter content, caution exchange capacity, potential of hydrogen, potassium, phosphorous, magnesium, calcium, or others.

14. The one or more non-transitory computer-readable media of claim 13 , storing additional instructions for: generating a plurality of subfield soil properties maps based on the information about the predicted subfield soil properties of the agricultural field; wherein the plurality of subfield soil properties maps includes one or more of: maps having different resolutions; maps depicting different soil properties, maps depicting different geographical regions, or maps depicting a geographical region at different points in time.

15. The one or more non-transitory computer-readable media of claim 11 , wherein the one or more machine learning models include one or more of: a Gaussian process regression (GPR) model, a random forest (RF) model, or any other non-parametric machine learning model;
wherein the one or more machine learning models have evolved from each other and include successive versions of an original machine learning model.

16. The one or more non-transitory computer-readable media of claim 11 , storing additional instructions for: identifying one or more management zones within the agricultural field based on the subfield soil property prediction for the agricultural field.

17. The one or more non-transitory computer-readable media of claim 11 , wherein the plurality of images of the satellite remote sensing data includes one or more of: an image in a visible optical domain, an image in a near-infrared optical domain, or an image in a short-wave infrared optical domain.

18. The one or more n

Assignees

• Climate Llc

Inventors

• Casas Angeles
• Yang Xiaoyuan
• Ward Steven

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• A01C21/007 Primary

  Determining fertilization requirements · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N20/20

  Ensemble learning · CPC title