US 12,481,875 B2
Method and apparatus for employing deep learning neural network to infer regenerative cover crop practices
Ernesto Brau, Newton, MA (US); R. Shane Bussmann, Cambridge, MA (US); and Ethan Sargent, Cambridge, MA (US)
Assigned to CIBO Technologies, Inc., St. Louis Park, MN (US)
Filed by CIBO Technologies, Inc., Cambridge, MA (US)
Filed on May 31, 2021, as Appl. No. 17/334,974.
Prior Publication US 2022/0383098 A1, Dec. 1, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G06N 3/08 (2023.01); G06F 16/53 (2019.01); G06N 3/045 (2023.01)
CPC G06N 3/08 (2013.01) [G06F 16/53 (2019.01); G06N 3/045 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method for determining crop-based agricultural management practices for use within a current growing year, the computer-implemented method comprising:
retrieving a first set of records from a historical cropland data layer database, wherein the first set of records corresponds to randomly sampled areas of a first geographic region taken over a first time period for a first number of years;
retrieving a second set of records from a historical imagery database, wherein the second set of records corresponds to the randomly sampled areas of the first geographic region, the first time period, and the first number of years;
employing the second set of records as inputs to train a first deep learning convolutional neural network to generate the first set of records and using parameters generated during training to configure a trained first deep learning convolutional neural network for execution;
configuring a second deep learning convolutional neural network using parameters corresponding to early layers of the trained first deep learning convolutional neural network;
retrieving a third set of records and a fourth set of records from an annotated imagery database, wherein the third set of records comprises unannotated image versions corresponding to a second geographic region, and wherein the fourth set of records comprises annotated image versions corresponding to the second geographic region, and wherein the annotated image versions comprise annotations indicative of management zones, and wherein the third and fourth sets of records correspond to a second time period for a second number of years;
employing the third set of records as inputs to train upper layers of the second deep learning convolutional neural network to generate the fourth set of records and using parameters generated during training to configure a trained second deep learning convolutional neural network for execution;
retrieving a fifth set of records from a current imagery database, wherein the fifth set of records comprises corresponds to a third geographic region, and wherein the fifth set of records corresponds to the second time period and the current growing year;
using the fifth set of records as inputs and executing the trained second deep learning convolutional neural network to generate predicted agricultural management zones for the current growing year; and
aggregating the fifth set of record into vegetative indices for parcels within the third geographic region, and processing the vegetative indices over the second time period for the current growing year to infer crop types, planting dates, harvest dates, and maturity for each of the predicted agricultural management zones as demarcated by boundaries of each of the parcels.
 
8. A non-transitory computer-readable storage medium storing instructions that, when executed by a computer, cause the computer to perform a method for determining crop-based agricultural management practices for use within a current growing year, the method comprising:
retrieving a first set of records from a historical cropland data layer database, wherein the first set of records corresponds to randomly sampled areas of a first geographic region taken over a first time period for a first number of years;
retrieving a second set of records from a historical imagery database, wherein the second set of records corresponds to the randomly sampled areas of the first geographic region, the first time period, and the first number of years;
employing the second set of records as inputs to train a first deep learning convolutional neural network to generate the first set of records and using parameters generated during training to configure a trained first deep learning convolutional neural network for execution;
configuring a second deep learning convolutional neural network using parameters corresponding to early layers of the trained first deep learning convolutional neural network;
retrieving a third set of records and a fourth set of records from an annotated imagery database, wherein the third set of records comprises unannotated image versions corresponding to a second geographic region, and wherein the fourth set of records comprises annotated image versions corresponding to the second geographic region, and wherein the annotated image versions comprise annotations indicative of management zones, and wherein the third and fourth sets of records correspond to a second time period for a second number of years;
employing the third set of records as inputs to train upper layers of the second deep learning convolutional neural network to generate the fourth set of records and using parameters generated during training to configure a trained second deep learning convolutional neural network for execution;
retrieving a fifth set of records from a current imagery database, wherein the fifth set of records comprises corresponds to a third geographic region, and wherein the fifth set of records corresponds to the second time period and the current growing year;
using the fifth set of records as inputs and executing the trained second deep learning convolutional neural network to generate predicted agricultural management zones for the current growing year; and
aggregating the fifth set of record into vegetative indices for parcels within the third geographic region, and processing the vegetative indices over the second time period for the current growing year to infer crop types, planting dates, harvest dates, and maturity for each of the predicted agricultural management zones as demarcated by boundaries of each of the parcels.
 
15. A computer program product for determining crop-based agricultural management practices for use within a current growing year, the computer program product comprising:
a non-transitory computer-readable medium having computer-readable program code stored thereon, the computer-readable program code comprising:
program instructions to retrieve a first set of records from a historical cropland data layer database, wherein the first set of records corresponds to randomly sampled areas of a first geographic region taken over a first time period for a first number of years;
program instructions to retrieve a second set of records from a historical imagery database, wherein the second set of records corresponds to the randomly sampled areas of the first geographic region, the first time period, and the first number of years;
program instructions to employ the second set of records as inputs to train a first deep learning convolutional neural network to generate the first set of records and to use parameters generated during training to configure a trained first deep learning convolutional neural network for execution;
program instructions to configure a second deep learning convolutional neural network using parameters corresponding to early layers of the trained first deep learning convolutional neural network;
program instructions to retrieve a third set of records and a fourth set of records from an annotated imagery database, wherein the third set of records comprises unannotated image versions corresponding to a second geographic region, and wherein the fourth set of records comprises annotated image versions corresponding to the second geographic region, and wherein the annotated image versions comprise annotations indicative of management zones, and wherein the third and fourth sets of records correspond to a second time period for a second number of years;
program instructions to employ the third set of records as inputs to train upper layers of the second deep learning convolutional neural network to generate the fourth set of records and to use parameters generated during training to configure a trained second deep learning convolutional neural network for execution;
program instructions to use the fifth set of records as inputs and to execute the trained second deep learning convolutional neural network to generate predicted agricultural management zones for the current growing year; and
program instructions to aggregate the fifth set of record into vegetative indices for parcels within the third geographic region, and to process the vegetative indices over the second time period for the current growing year to infer crop types, planting dates, harvest dates, and maturity for each of the predicted agricultural management zones as demarcated by boundaries of each of the parcels.