System and method for hyperspectral image processing to identify object

1. A system comprising:

a memory; and

at least one processor to:

acquire an actual hyperspectral image of a food item by an imaging device;

receive measurements of the food item from one or more food measurement devices, wherein the measurements indicate chemical and physical properties of the food item that represent different quality characteristics of the food item;

generate, based on applying one or more machine learning models to the actual hyperspectral image of the food item, a plurality of synthetic hyperspectral images of the food item, wherein the one or more machine learning models were trained to correlate the actual hyperspectral image of the food item with the measurements of the food item in different perspectives and visual angles; and

return the plurality of synthetic hyperspectral images of the food item to be used to determine particular characteristics of the food item and determine a value for at least one quality parameter for the food item.

2. The system of claim 1 , wherein the at least one processor configured to generate the plurality of synthetic hyperspectral images comprises replicating first color pixels and second color pixels from the actual hyperspectral image of the food item into different positions in one or more of the plurality of synthetic hyperspectral images, wherein each of the plurality of synthetic hyperspectral images represent a different stage of ripeness for the food item.

3. The system of claim 1 , wherein the at least one processor configured to generate the plurality of synthetic hyperspectral images comprises augmenting the actual hyperspectral image of the food item based on adding a first color to pixels of the actual hyperspectral image, wherein the augmented hyperspectral image represents a change in ripeness of the food item.

4. The system of claim 1 , wherein the at least one processor configured to generate the plurality of synthetic hyperspectral images comprises augmenting the actual hyperspectral image of the food item based on inserting spectral information into the actual hyperspectral image.

5. The system of claim 1 , wherein the at least one processor configured to generate the plurality of synthetic hyperspectral images comprises augmenting the actual hyperspectral image of the food item based on inserting a foreign object into the actual hyperspectral image at one or more positions and orientations.

6. The system of claim 5 , wherein the foreign object is at least one of plastic, a cable tie, paper, a metal hook, a cigarette butt, cardboard, and textile.

7. The system of claim 1 , wherein the at least one processor is further configured to generate a three-dimensional (3D) model of the food item based on the actual hyperspectral image of the food item and the measurements of the food item.

8. The system of claim 1 , wherein the at least one processor configured to generate the plurality of synthetic hyperspectral images comprises augmenting the actual hyperspectral image of the food item based on adding one or more bruises on the food item in the actual hyperspectral image, wherein the food item is at least one of a citrus fruit.

9. The system of claim 1 , wherein the at least one processor configured to generate the plurality of synthetic hyperspectral images comprises augmenting the actual hyperspectral image of the food item based on adding a green color to the food item, wherein the food item is a banana, wherein the actual hyperspectral image depicts the banana in a yellow color, and wherein the yellow color indicates that the banana has a firmness below a threshold value and the green color indicates that the banana has a firmness above the threshold value.

10. The system of claim 1 , wherein the at least one processor configured to generate the plurality of synthetic hyperspectral images comprises augmenting the actual hyperspectral image of the food item based on adding a cold injury to the food item in the actual hyperspectral image, wherein the food item is a fruit.

11. The system of claim 1 , wherein the at least one processor configured to generate the plurality of synthetic hyperspectral images comprises augmenting the actual hyperspectral image of the food item based on adding sour skin disease to the food item in the actual hyperspectral image, wherein the food item is an onion.

12. The system of claim 1 , wherein the at least one processor is further configured to:

train an object model using the plurality of synthetic hyperspectral images of the food item and the measurements associated with the food item; and

determine, based on applying the object model to the actual hyperspectral image of the food item, the particular characteristics of the food item.

13. The system of claim 1 , wherein the imaging device is a hyperspectral imaging device and the system further comprises at least one illumination device that illuminates the food item and a conveyor belt that moves the food item past the hyperspectral imaging device as the hyperspectral imaging device captures the actual hyperspectral image of the food item.

14. The system of claim 1 , wherein the at least one processor is further configured further to determine the particular characteristics of the food item based on the plurality of synthetic hyperspectral images of the food item.

15. A method for generating a synthetic training set of images of a food item, the method comprising:

acquiring an actual hyperspectral image of a food item by an imaging device;

receiving measurements of the food item from one or more food measurement devices, wherein the measurements indicate chemical and physical properties of the food item that represent different quality characteristics of the food item;

generating, based on applying one or more machine learning models to the actual hyperspectral image of the food item, a plurality of synthetic hyperspectral images of the food item, wherein the one or more machine learning models were trained to correlate the actual hyperspectral image of the food item with the measurements of the food item in different perspectives and visual angles; and

returning the plurality of synthetic hyperspectral images of the food item to be used to determine particular characteristics of the food item and determine a value for at least one quality parameter for the food item.

16. The method of claim 15 , wherein generating the plurality of synthetic hyperspectral images comprises replicating first color pixels and second color pixels from the actual hyperspectral image of the food item into different positions in one or more of the plurality of synthetic hyperspectral images, wherein each of the plurality of synthetic hyperspectral images represent a different stage of ripeness for the food item.

17. The method of claim 15 , wherein generating the plurality of synthetic hyperspectral images comprises augmenting the actual hyperspectral image of the food item based on adding a first color to pixels of the actual hyperspectral image, wherein the augmented hyperspectral image represents a change in ripeness of the food item.

18. The method of claim 15 , wherein generating the plurality of synthetic hyperspectral images comprises augmenting the actual hyperspectral image of the food item based on inserting spectral information into the actual hyperspectral image.

19. The method of claim 15 , wherein generating the plurality of synthetic hyperspectral images comprises augmenting the actual hyperspectral image of the food item based on inserting a foreign object into the actual hyperspectral image at one or more positions and orientations.

20. The method of claim 19 , wherein the foreign object is at least one of plastic, a cable tie, paper, a metal hook, a cigarette butt, cardboard, and textile.