Discovering farming practices

Patent metadata

Field	Value
Publication number	US-12456060-B2
Application number	US-202016950266-A
Country	US
Kind code	B2
Filing date	Nov 17, 2020
Priority date	Nov 17, 2020
Publication date	Oct 28, 2025
Grant date	Oct 28, 2025

Abstract

One embodiment provides a method, including: training a machine-learning model to produce customized farming practices specific to a farm to increase crop yield; wherein the training includes obtaining, from remote sensed data, (i) information corresponding to a crop of each of a plurality of farms and (ii) information corresponding to farming practices of each of the plurality of farms; wherein the training further includes detecting, from the remote sensed data, geographical features and farming characteristics of each of the plurality of farms; wherein the machine-learning model identifies from relationships between (iii) crop information and farming practices and (iv) geographical features and farming characteristics; and discovering, for a specific farm in an identified geographical location, utilizing the trained machine-learning model, and from farm-specific remote-sensed data, farming practices.

First claim

What is claimed is:

1. A computer implemented method, comprising:
training a machine-learning model to produce customized farming practices specific to a farm to increase crop yield;
wherein the training comprises obtaining, from remote sensed data, (i) information corresponding to a crop of each of a plurality of farms and (ii) information corresponding to farming practices of each of the plurality of farms, wherein the farming practices further comprises different farming techniques;
wherein the training further comprises detecting, from the remote sensed data, geographical features and farming characteristics of each of the plurality of farms;
wherein the machine-learning model identifies relationships between (iii) the obtained crop information and farming practices information and (iv) the detected geographical features and farming characteristics;
automatically discovering, for a specific farm in an identified geographical location, utilizing the trained machine-learning model, and from farm-specific remote-sensed data including the farming characteristics, current farming practices; and
based on the discovered current farming practices, determining, utilizing the trained machine-learning model, the customized farming practices including changes and improvements to the discovered current farming practices, wherein determining the customized farming practices further comprises correlating the crop yield with the discovered current farming practices, and recommending the customized farming practices for improving the crop yield.

2. The computer implemented method of claim 1 , wherein the remote-sensed data comprises at least one of active remote sensing and passive remote sensing.

3. The computer implemented method of claim 1 , wherein the obtaining from the remote sensed data, information corresponding to farming practices comprises obtaining metadata identifying historical crop data.

4. The computer implemented method of claim 1 , wherein the detecting, from the remote sensed data, geographical features and farming characteristics comprises analyzing synthetic aperture radar for backscatter signals, wherein the backscatter signals provide an indication of the geographical features and farming characteristics.

5. The computer implemented method of claim 4 , wherein the geographical features and farming characteristics are indicated via the backscatter signals by constructing at least one of: a scattering matrix, a coherence matrix, and a covariance matrix.

6. The computer implemented method of claim 1 , wherein the identifying relationships comprises clustering the obtained crop information and farming practices and the detected geographical features and farming characteristics.

7. The computer implemented method of claim 1 , wherein the farm-specific remote-sensed data comprises remote images of the specific farm and wherein the discovering comprises performing a pixel to pixel comparison on the remote images to identify crop locations on the specific farm.

8. The computer implemented method of claim 1 , wherein the discovering comprises clustering rows of crops identified from the farm-specific remote-sensed data and identifying a proximity of clusters to identify row spacing on the specific farm.

9. The computer implemented method of claim 1 , wherein the farm-specific remote-sensed data comprises synthetic aperture radar images and wherein the discovering comprises detecting, utilizing synthetic aperture radar interferometry on the synthetic aperture radar images, surface changes to identify hilling activity on the specific farm.

10. The computer implemented method of claim 1 , wherein the farming practices comprises at least one of: row-to-row spacing, plant-to-plant spacing, plant density, hilling activity, frequency of irrigation, de-weeding, row orientation, and gap filling activity.

11. An apparatus, comprising:
at least one processor; and a computer readable storage medium having a computer readable program code embodied therewith and executable by the at least one processor;
wherein the computer readable program code is configured to train a machine-learning model to produce customized farming practices specific to a farm to increase crop yield;
wherein the training comprises obtaining, from remote sensed data, (i) information corresponding to a crop of each of a plurality of farms and (ii) information corresponding to farming practices of each of the plurality of farms, wherein the farming practices further comprises different farming techniques;
wherein the training further comprises detecting, from the remote sensed data, geographical features and farming characteristics of each of the plurality of farms;
wherein the machine-learning model identifies relationships between (iii) the obtained crop information and farming practices information and (iv) the detected geographical features and farming characteristics;
automatically discovering, for a specific farm in an identified geographical location, utilizing the trained machine-learning model, and from farm-specific remote-sensed data including the farming characteristics, current farming practices; and
based on the discovered current farming practices, determining, utilizing the trained machine-learning model, the customized farming practices including changes and improvements to the discovered current farming practices, wherein determining the customized farming practices further comprises correlating the crop yield with the discovered current farming practices, and recommending the customized farming practices for improving the crop yield.

12. A computer program product, comprising:
a computer readable storage medium having computer readable program code embodied therewith, the computer readable program code executable by the processor;
wherein the computer readable program code is configured to train a machine-learning model to produce customized farming practices specific to a farm to increase crop yield;
wherein the training comprises obtaining, from remote sensed data, (i) information corresponding to a crop of each of a plurality of farms and (ii) information corresponding to farming practices of each of the plurality of farms, wherein the farming practices further comprises different farming techniques;
wherein the training further comprises detecting, from the remote sensed data, geographical features and farming characteristics of each of the plurality of farms;
wherein the machine-learning model identifies relationships between (iii) the obtained crop information and farming practices information and (iv) the detected geographical features and farming characteristics;
automatically discovering, for a specific farm in an identified geographical location, utilizing the trained machine-learning model, and from farm-specific remote-sensed data including the farming characteristics, current farming practices; and
based on the discovered current farming practices, determining, utilizing the trained machine-learning model, the customized farming practices including changes and improvements to the discovered current farming practices, wherein determining the customized farming practices further comprises correlating the crop yield with the discovered current farming practices, and recommending the customized farming practices for improving the crop yield.

13. The computer program product of claim 12 , wherein the remote-sensed data comprises at least one of active remote sensing and passive remote sensing.

14. The computer program product of claim 12 , wherein the obtaining from the remote sensed data, information corresponding to farming practices compri

Assignees

• IBM

Inventors

• MARVANIYA Smitkumar Narotambhai
• Devi Umamaheswari
• Godbole Shantanu R

Classifications

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06V20/188

  Vegetation · CPC title

• G06V10/758

  Involving statistics of pixels or of feature values, e.g. histogram matching · CPC title

• G06T2207/30188

  Vegetation; Agriculture · CPC title

• A01B79/005 Primary

  Precision agriculture · CPC title