Automated optimization of agricultural treatments based on raster image data system
Systems and methods to collect, combine, store, and process remote sensing and imaging information as well for agricultural area as a raster image data system for handling both remotely-sensed aerial image data and land-based sensor data, as well as previous agricultural treatments, in order to optimize future agricultural treatments to variable resolutions of the area of vegetation corresponding to that represented by a pixel in a raster image data system of the area.
1. A system for optimizing the application of one or more treatments to an agricultural area, the system comprising:
a learning engine comprising instructions that, when implemented by a processor operably coupled to memory, cause the processor to:
receive a plurality of training raster images each having a plurality of pixels, the plurality of training raster images including a set of parameters corresponding to a set of intersection points of each training raster image at each relative pixel location,
learn a relationship between the set of parameters and at least one biological stress, and
output a probability of the at least one biological stress for each of the set of intersection points in the training raster images as a trained learning engine output;
a first remote data capture device configured to capture data related to the agricultural area in a first raster image;
a second remote data capture device configured to provide data related to the agricultural area in a second raster image;
a data cube storage configured to store the first raster image and the second raster image in an agricultural raster data cube, the agricultural raster data cube including a set of detected parameters corresponding to a set of intersection points of the first raster image and the second raster image at each relative pixel location; and
an application engine comprising instructions that, when implemented by a processor operably coupled to memory, cause the processor to:
receive the trained learning engine output,
determine a probability of disease based on the set of detected parameters in the agricultural raster data cube for each pixel location and the probability of the at least one biological stress for each of the set of intersection points in the trained learning engine output, and
output a proposed treatment raster specifying an amount of at least one of the treatments for at least some of the pixel locations.
2. The system of claim 1 , further comprising a precision applicator configured to apply the at least one of the treatments to the agricultural area according to the proposed treatment raster as provided in a format specific to the precision applicator for each of the at least some of the pixel locations.
3. The system of claim 2 , wherein an actual treatment raster is generated based on the application of the at least one of the treatments to the agricultural area by the precision applicator and added to the agricultural raster data cube.
4. The system of claim 1 , wherein the first raster image and the second raster image are georeferenced to the same physical location.
5. The system of claim 4 , wherein the data cube storage further comprises a second agricultural raster data cube proximate the georeferenced location of the agricultural raster data cube, and wherein the agricultural raster data cube comprises raster images at a first precision and the second agricultural raster data cube comprises raster images at a second precision, the second precision being more accurate than the first precision.
6. The system of claim 1 , wherein the first remote data capture device is positioned on an unmanned aerial vehicle.
7. The system of claim 1 , wherein the second remote data capture device is a ground-based remote sensor.
8. The system of claim 1 , wherein the first remote data capture device is configured to capture the same agricultural area in a first image at a first time and a second image at a subsequent second time and the learning engine is further configured to calculate a stress vector that represents a prediction of the at least one biological stress based on the first image and the second image.
9. The system of claim 8 , wherein the proposed treatment raster pre-treats the at least one biological stress with the proposed treatment raster.
10. The system of claim 1 , wherein the first raster image and the second raster image are stored in the agricultural raster data cube as delta values.
11. A method for optimizing the application of one or more treatments to an agricultural area, the method comprising:
training a machine learning network with a plurality of training raster images each having a plurality of pixels, the plurality of training raster images including a set of parameters corresponding to a set of intersection points of each training raster image at each relative pixel location,
training the machine learning network with a plurality of known diseases, including learning a relationship between the set of parameters and at least one biological stress, and output a probability of the at least one biological stress for each of the set of intersection points in the training raster images as a trained machine learning network output;
capturing a plurality of raster images using one or more image capture devices;
providing a plurality of raster images based on one or more remote sensing devices;
storing the plurality of raster images as an agricultural raster data cube in a data cube storage, the agricultural raster data cube including a set of detected parameters corresponding to a set of intersection points of the plurality of raster images at each relative pixel location;
determining a probability of disease for each pixel location in the agricultural raster data cube based on the set of detected parameters and the probability of the at least one biological stress for each of the set of intersection points in the trained machine learning network output, and
outputting a proposed treatment raster specifying an amount of treatment material for each pixel in a precision applicator format.
12. The method of claim 11 , further comprising applying the one or more treatments to the agricultural area according to the proposed treatment raster.
13. The method of claim 12 , further comprising:
generating an actual treatment raster based on an application of the one or more treatments, and
adding the actual treatment raster to the agricultural raster data cube.
14. The method of claim 11 , wherein the plurality of raster images are georeferenced to the same physical location.
15. The method of claim 14 , further comprising storing a second agricultural raster data cube proximate the georeferenced location of the agricultural raster data cube, and wherein the agricultural raster data cube comprises raster images at a first precision and the second agricultural raster data cube comprises raster images at a second precision, the second precision being more accurate than the first precision.
16. The method of claim 11 , wherein the one or more image capture devices is positioned on an unmanned aerial vehicle.
17. The method of claim 11 , wherein the one or more remote sensing devices is a ground-based remote sensor.
18. The method of claim 11 , wherein capturing a plurality of raster images using one or more image capture devices includes:
capturing the same agricultural area in a first image at a first time and a second image at a subsequent second time; and
calculating a stress vector that represents a prediction of at least one of the plurality of known diseases based on the first image and the second image.
19. The method of claim 18 , further comprising pre-treating the predicted at least one of the plurality of known diseases with the proposed treatment raster.
20. The method of claim 11 , wherein storing the plurality of raster images as an agricultural raster data cube further comprises storing the first raster image and the second raster image as delta values.