Methods for unit load device (ULD) localization

1. A method for localizing a unit load device (ULD), comprising:

capturing a set of image data featuring the ULD;

cropping the set of image data based on an expected location of the ULD within the set of image data to generate a cropped image, the cropped image featuring a portion of the ULD;

determining one or more candidate edges of the portion of the ULD within the cropped image; and

identifying one or more edges of the portion of the ULD from the one or more candidate edges, wherein each of the one or more edges represents a boundary of the portion of the ULD.

2. The method of claim 1 , wherein the set of image data featuring the ULD comprises (i) a three-dimensional depth image and (ii) a grayscale image that is depth-aligned with the three-dimensional depth image.

3. The method of claim 1 , wherein the set of image data featuring the ULD comprises (i) a three-dimensional depth image and (ii) a red-green-blue (RGB) image, and wherein the method further comprises:

aligning the RGB image with the three-dimensional depth image.

4. The method of claim 1 , wherein the expected location includes (i) an expected height of the ULD or (ii) an expected width of the ULD.

5. The method of claim 1 , further comprising:

performing (i) a height-based cropping, (ii) a width-based cropping, or (iii) a depth-based cropping to crop the set of image data to generate the cropped image.

6. The method of claim 1 , further comprising:

applying (i) a thresholding technique or (ii) a canny edge detection filter to the cropped image to determine the one or more candidate edges of the portion of the ULD within the cropped image.

7. The method of claim 1 , further comprising:

applying a set of heuristic principles to identify the one or more edges of the portion of the ULD from the one or more candidate edges, the set of heuristic principles including (i) a number of boundaries for the portion of the ULD, (ii) an orientation of each boundary of the portion of the ULD, or (iii) a perimeter measurement of the boundaries of the portion of the ULD.

8. The method of claim 1 , further comprising:

training a machine learning model using (i) a plurality of cropped images, each cropped image featuring a portion of a respective ULD, (ii) one or more candidate edges of the portion of the respective ULD within each cropped image, and (iii) one or more edges of the portion of the respective ULD from the one or more candidate edges, wherein each of the one or more edges represents a boundary of the portion of the respective ULD; and

applying the machine learning model to the cropped image featuring the portion of the ULD to identify the one or more edges of the portion of the ULD from the one or more candidate edges.

9. The method of claim 1 , wherein the set of image data is a first set of image data, and the method further comprises:

(a) capturing a second set of image data featuring the ULD a duration D after capturing the first set of image data;

(b) designating the second set of image data as a current set of image data;

(c) cropping the current set of image data to generate a current cropped image, the current cropped image featuring a current portion of the ULD;

(d) determining one or more current candidate edges of the current portion of the ULD within the current cropped image;

(e) identifying one or more current edges of the current portion of the ULD from the one or more current candidate edges, wherein each of the one or more current edges represents a current boundary of the current portion of the ULD;

(e) designating the current set of image data as a prior set of image data;

capturing a subsequent set of image data featuring the ULD the duration D after capturing the prior set of image data;

(g) designating the subsequent set of image data as the current set of image data; and

(h) iteratively performing steps (c)-(h) until the current set of image data does not feature the ULD.

10. A system for localizing a unit load device (ULD), comprising:

a housing;

an imaging assembly at least partially within the housing and configured to capture a set of image data featuring the ULD;

one or more processors; and

a non-transitory computer-readable memory coupled to the imaging assembly and the one or more processors, the memory storing instructions thereon that, when executed by the one or more processors, cause the one or more processors to:

crop the set of image data based on an expected location of the ULD within the set of image data to generate a cropped image, the cropped image featuring a portion of the ULD;

determine one or more candidate edges of the portion of the ULD within the cropped image; and

identify one or more edges of the portion of the ULD from the one or more candidate edges, wherein each of the one or more edges represents a boundary of the portion of the ULD.

11. The system of claim 10 , wherein the set of image data featuring the ULD comprises (i) a three-dimensional depth image and (ii) a grayscale image that is depth-aligned with the three-dimensional depth image.

12. The system of claim 10 , wherein the set of image data featuring the ULD comprises (i) a three-dimensional depth image and (ii) a red-green-blue (RGB) image, and wherein the instructions further cause the one or more processors to:

align the RGB image with the three-dimensional depth image.

13. The system of claim 10 , wherein

the expected location includes (i) an expected height of the ULD or (ii) an expected width of the ULD.

14. The system of claim 10 , wherein the instructions further cause the one or more processors to:

perform (i) a height-based cropping, (ii) a width-based cropping, or (iii) a depth-based cropping to crop the set of image data to generate the cropped image.

15. The system of claim 10 , wherein the instructions further cause the one or more processors to:

apply (i) a thresholding technique or (ii) a canny edge detection filter to the cropped image to determine the one or more candidate edges of the portion of the ULD within the cropped image.

16. The system of claim 10 , wherein the instructions further cause the one or more processors to:

apply a set of heuristic principles to identify the one or more edges of the portion of the ULD from the one or more candidate edges, the set of heuristic principles including (i) a number of boundaries for the portion of the ULD, (ii) an orientation of each boundary of the portion of the ULD, or (iii) a perimeter measurement of the boundaries of the portion of the ULD.

17. The system of claim 10 , wherein the instructions further cause the one or more processors to:

train a machine learning model using (i) a plurality of cropped images, each cropped image featuring a portion of a respective ULD, (ii) one or more candidate edges of the portion of the respective ULD within each cropped image, and (iii) one or more edges of the portion of the respective ULD from the one or more candidate edges, wherein each of the one or more edges represents a boundary of the portion of the respective ULD; and

apply the machine learning model to the cropped image featuring the portion of the ULD to identify the one or more edges of the portion of the ULD from the one or more candidate edges.

18. The system of claim 10 , wherein the set of image data is a first set of image data, and wherein the instructions further cause the one or more processors to:

(a) receive a second set of image data featuring the ULD a duration D after receiving the first set of image data;

(b) designating the second set of image data as a current set of image data;

(c) cropping the current set of image data to generate a current cropped image, the current cropped image featuring a current portion of the ULD;

(d) determining one or more current candidate edges of the current portion of the ULD within the current cropped image;

(e) identifying one or more current edges of the current portion of the ULD from the one or more current candidate edges, wherein each of the one or more current edges represents a current boundary of the current portion of the ULD;

(e) designating the current set of image data as a prior set of image data;

receiving a subsequent set of image data featuring the ULD the duration D after receiving the prior set of image data;

(g) designating the subsequent set of image data as the current set of image data; and

(h) iteratively performing steps (c)-(h) until the current set of image data does not feature the ULD.