Method and apparatus for automated plant necrosis

1. A method comprising:

capturing a first image of a first section of a field of plants with a camera physically attached to a machine travelling in a first direction through the field, the first image comprising an array of pixels;

segmenting the first image into a plurality of regions, the plurality of regions segmented based on differences in depth between the array of pixels of the first image;

identifying a plurality of points of interest within one of the plurality of regions;

classifying by a computer, a first point of the plurality of points of interest as a first unique feature of a first distinct plant growing in the field;

classifying, by the computer, a second point of the plurality of points of interest as a second unique feature of a second distinct plant growing in the field, the second distinct plant at least partially overlapping the first distinct plant;

selecting, by the computer, the first distinct plant for removal;

sending instructions for removal of the first distinct plant from the computer to a plant removal mechanism that is also physically attached to the machine, the instructions sent before the plant removal mechanism passes by the first distinct plant along the first direction; and

activating the plant removal mechanism responsive to the instructions as the plant removal mechanism passes by the first distinct plant along the first direction.

2. The method of claim 1 wherein the first unique feature of the first plant growing in the field is a stalk of a plant.

3. The method of claim 1 wherein the identifying the plurality of points of interest within one of the plurality of regions comprises performing, based on the array of pixels associated within the one of the plurality of regions, at least one from the group consisting of:

a feature detection gradient analysis,

a hue analysis,

a brightness analysis,

a shape analysis, and

a blob analysis.

4. The method of claim 1 wherein the classifying the first point of the plurality of the points of interest as the first unique feature of the firs 6 t distinct plant growing in the field comprises applying the plurality of the points of interest to a trained machine learning model to obtain a numerical confidence level for each point of the plurality of points of interest representing a likelihood that the first point of interest represents the first unique feature of the plant.

5. The method of claim 1 further comprising:

capturing a sequence of images, the first image being a part of the sequence;

segmenting the sequence of images into regions;

identifying the plurality of points of interests in one of the plurality of regions in each of the images in the sequence; and

classifying a subset of the points of interest identified in a plurality of the images in the sequence together as being the first unique feature of the first distinct plant growing in the field, the first unique feature identifying the first distinct plant growing in the field across multiple of the images in the sequence.

6. The method of claim 5

wherein the classifying the subset of the points of interest comprises adjusting locations of the subset of the points of interest within their respective regions and images of the sequence by accounting for changes in position of the machine between images of the sequence of images.

7. The method of claim 5 further comprising:

creating a virtual map of plants that places the first distinct plant at a position within the virtual map based on the subset of the points of interest in the images in the sequence used to classify the first unique feature of the first distinct plant; and

updating the virtual map to include additional plants placed at positions within the virtual map based on other points of interests in the images in the sequence used to classify the first unique features of the additional plants.

8. The method of claim 7 wherein

tracking an absolute geographic position of the machine as a function of time during the capture of the images of the sequence; and

wherein placing the first distinct plants at positions within the virtual map is further based on the absolute geographic position.

9. A non-transitory computer readable medium storing instructions, the instructions when executed cause a processor physically attached to a machine travelling in a first direction through a field of plants to:

capture a first image of a first section of a field of plants with a camera physically attached to a machine travelling in a first direction through the field, the first image comprising an array of pixels;

segment the first image into a plurality of regions, the plurality of regions segmented based on differences in depth between the array of pixels of the first image;

identify a plurality of points of interest within one of the plurality of regions;

classify by a computer, a first point of the plurality of points of interest as a first unique feature of a first distinct plant growing in the field;

classify, by the computer, a second point of the plurality of points of interest as a second unique feature of a second distinct plant growing in the field, the second distinct plant at least partially overlapping the first distinct plant;

select, by the computer, the first distinct plant for removal;

send instructions for removal of the first distinct plant from the computer to a plant removal mechanism that is also physically attached to the machine, the instructions sent before the plant removal mechanism passes by the first distinct plant along the first direction; and

activate the plant removal mechanism responsive to the instructions as the plant removal mechanism passes by the first distinct plant along the first direction.

10. The non-transitory computer readable medium of claim 9 wherein the first unique feature of the first plant growing in the field is at least one from the group consisting of: a stalk of a plant, a point of contact between the plant and the ground, and a plant center.

11. The non-transitory computer readable medium of claim 9 wherein the identifying the plurality of points of interest within one of the plurality of regions comprises performing, based on the array of pixels associated within the one of the plurality of regions, at least one from the group consisting of:

a feature detection gradient analysis,

a hue analysis,

a brightness analysis,

a shape analysis, and

a blob analysis.

12. The non-transitory computer readable medium of claim 9 wherein the classifying the first point of the plurality of the points of interest as the first unique feature of the first distinct plant growing in the field comprises applying the plurality of the points of interest to a trained machine learning model to obtain a numerical confidence level for each point of the plurality of points of interest representing a likelihood that the first point of interest represents the first unique feature of the plant.

13. The non-transitory computer readable medium of claim 9 wherein the instructions further cause the processor to:

capture a sequence of images, the first image being a part of the sequence;

segment the sequence of images into regions;

identify the plurality of points of interests in one of the plurality of regions in each of the images in the sequence; and

classify a subset of the points of interest identified in a plurality of the images in the sequence together as being the first unique feature of the first distinct plant growing in the field, the first unique feature identifying the first distinct plant growing in the field which is depicted across multiple of the images in the sequence.

14. The non-transitory computer readable medium of claim 13

wherein the classifying the subset of the points of interest comprises adjusting locations of the subset of the points of interest within their respective regions and images of the sequence by accounting for changes in position of the machine between images of the sequence of images.

15. The non-transitory computer readable medium of claim 13 wherein the instructions further cause the processor to:

create a virtual map of plants that places the first distinct plant at a position within the virtual map based on the subset of the points of interest in the images in the sequence used to classify the first unique feature of the first distinct plant; and

update the virtual map to include additional plants placed at positions within the virtual map based on other points of interests in the images in the sequence used to classify the first unique features of the additional plants.

16. The non-transitory computer readable medium of claim 15 wherein the instructions further cause the processor to:

track an absolute geographic position of the machine as a function of time during the capture of the images of the sequence; and

wherein placing the first distinct plants at positions within the virtual map is further based on the absolute geographic position.

17. The method of claim 1 wherein the selecting the first distinct plant for removal comprises

sending the first image to a user interface device from the machine, the user interface device configured to display the first and the second distinct plants;

providing an option to a user via the user interface device to select the first distinct plant for removal;

receiving the selection of the first distinct plant for removal from the user via the interface device;

sending instructions for removal of the first distinct plant for removal to the machine.

18. The method of claim 1 wherein selecting the first distinct plant for removal comprises

classifying the first distinct plant with a type, the type acting as an identifier for at least one of a plant species, genus, or family;

selecting the first distinct plant for removal based on the type.

19. The method of claim 1 further comprising:

capturing a second image of the first section of the field of plants with a second camera physically attached to the machine, the second image captured after the first image;

verifying with the second image that the first distinct plant was acted upon by the plant removal mechanism.

20. The method of claim 1 wherein the plant removal mechanism further comprises a spray nozzle for dispensing a mechanism for plant necrosis.

21. The non-transitory computer readable medium of claim 9 wherein the instructions to select the first distinct plant for removal further cause the processor to:

send the first image to a user interface device from the machine, the user interface device configured to display the first and the second distinct plants;

provide an option to a user via the user interface device to select the first distinct plant for removal;

receive the selection of the first distinct plant for removal from the user via the interface device;

send instructions for removal of the first distinct plant for removal to the machine.

22. The non-transitory computer readable medium of claim 9 wherein the instructions further cause the processor to:

classify the identified plants with a type, the type acting as an identifier for at least one of a plant species, genus, or family;

select at least one plant for removal based on the type.

23. The non-transitory computer readable medium of claim 9 wherein the instructions further cause the processor to:

capture a second image of the first section of the field of plants with a second camera physically attached to the machine, the second image captured after the first image;

verify with the second image that the first distinct plant was acted upon by the plant removal mechanism.

24. The non-transitory computer readable medium of claim 9 wherein the plant removal mechanism comprises a spray nozzle for dispensing a mechanism for plant necrosis.