Method for detecting trailer status using combined 3D algorithms and 2D machine learning models
Methods for determining a trailer status are disclosed herein. An example method includes capturing a three-dimensional image and a two-dimensional image. The three-dimensional image may comprise three-dimensional image data, and the two-dimensional image may comprise two-dimensional image data. The example method may further include determining a first trailer status based on the three-dimensional image data, and determining a second trailer status based on the two-dimensional image data. The example method may further include comparing the first trailer status to the second trailer status to determine a final trailer status.
1. A method for determining a trailer status, comprising:
capturing a three-dimensional image and a two-dimensional image, the three-dimensional image comprising three-dimensional image data, and the two-dimensional image comprising two-dimensional image data;
determining a first trailer status based on the three-dimensional image data;
determining a second trailer status based on the two-dimensional image data; and
comparing the first trailer status to the second trailer status to determine a final trailer status
wherein determining the second trailer status further comprises:
training a first machine learning model based on (i) a first set of prior two-dimensional image data and (ii) a first set of corresponding trailer statuses, and a second machine learning model based on (i) a second set of prior two-dimensional image data and (ii) a second set of corresponding trailer statuses;
applying the first machine learning model to the two-dimensional image data to generate a preliminary trailer status, wherein the preliminary trailer status indicates one of (i) a closed trailer door, (ii) an other trailer status, or (iii) an unknown trailer status; and
in response to generating a preliminary trailer status indicating the other trailer status, applying the second machine learning model to the two-dimensional image data to generate the second trailer status.
2. The method of claim 1 , further comprising:
in response to generating the preliminary trailer status indicating the other trailer status, cropping the two-dimensional image data; and
applying the second machine learning model to the cropped two-dimensional image data to generate the second trailer status.
3. The method of claim 1 , wherein the first set of corresponding trailer statuses includes indications that a respective image in the first set of prior two-dimensional image data represents at least one of (i) the closed trailer door, (ii) the other trailer status, or (iii) the unknown trailer status;
wherein the second set of prior two-dimensional image data is a set of prior cropped two-dimensional data; and
wherein the second set of corresponding trailer statuses includes indications that a respective image in the set of prior cropped two-dimensional image data represents at least one of (i) an ajar trailer door, (ii) an open trailer door, or (iii) a parking lot.
4. The method of claim 1 , wherein comparing the first trailer status to the second trailer status to determine the final trailer status further comprises determining whether the first trailer status is substantially similar to the second trailer status; and
responsive to determining that the first trailer status is substantially similar to the second trailer status, determining the final trailer status based on either the first trailer status or the second trailer status; and
responsive to determining that the first trailer status is not substantially similar to the second trailer status, determining the final trailer status based on a set of tested values indicating which of the first trailer status and the second trailer status is more accurate.
5. An apparatus for determining a trailer status, comprising:
a housing;
an imaging assembly at least partially within the housing and configured to capture a three-dimensional image and a two-dimensional image, the three-dimensional image comprising three-dimensional image data, and the two-dimensional image comprising two-dimensional image data; and
a controller communicatively coupled to the imaging assembly, the controller having a processor and a memory, the memory storing instructions that, when executed by the processor, cause the controller to:
determine a first trailer status based on the three-dimensional image data;
determine a second trailer status based on the two-dimensional image data; and
compare the first trailer status to the second trailer status to determine a final trailer status
wherein the instructions further cause the controller to:
train a first machine learning model based on (i) a first set of prior two-dimensional image data and (ii) a first set of corresponding trailer statuses, and a second machine learning model based on (i) a second set of prior two-dimensional image data and (ii) a second set of corresponding trailer statuses;
apply the first machine learning model to the two-dimensional image data to generate a preliminary trailer status, wherein the preliminary trailer status indicates one of (i) a closed trailer door, (ii) an other trailer status, or (iii) an unknown trailer status; and
in response to generating a preliminary trailer status indicating the other trailer status, apply the second machine learning model to the two-dimensional image data to generate the second trailer status.
6. The apparatus of claim 5 , wherein the instructions further cause the controller to:
in response to generating the preliminary trailer status indicating the other trailer status, crop the two-dimensional image data; and
apply the second machine learning model to the cropped two-dimensional image data to generate the second trailer status.
7. The apparatus of claim 5 , wherein the first set of corresponding trailer statuses includes indications that a respective image in the first set of prior two-dimensional image data represents at least one of (i) the closed trailer door, (ii) the other trailer status, or (iii) the unknown trailer status;
wherein the second set of prior two-dimensional image data is a set of prior cropped two-dimensional data; and
wherein the second set of corresponding trailer statuses includes indications that a respective image in the set of prior cropped two-dimensional image data represents at least one of (i) an ajar trailer door, (ii) an open trailer door, or (iii) a parking lot.
8. The apparatus of claim 5 , wherein the instructions further cause the controller to:
compare the first trailer status to the second trailer status to determine the final trailer status by determining whether the first trailer status is substantially similar to the second trailer status;
responsive to determining that the first trailer status is substantially similar to the second trailer status, determine the final trailer status based on either the first trailer status or the second trailer status; and
responsive to determining that the first trailer status is not substantially similar to the second trailer status, determine the final trailer status based on a set of tested values indicating which of the first trailer status and the second trailer status is more accurate.
9. A system for determining a trailer status, comprising:
a user interface;
a trailer monitoring unit (TMU) mounted proximate a loading bay and communicatively connected with the user interface, the TMU including:
a housing; and
an imaging assembly at least partially within the housing and configured to capture a three-dimensional image and a two-dimensional image, the three-dimensional image comprising three-dimensional image data, and the two-dimensional image comprising two-dimensional image data;
wherein the TMU is configured to:
determine a first trailer status based on the three-dimensional image data;
determine a second trailer status based on the two-dimensional image data; and
compare the first trailer status to the second trailer status to determine a final trailer status
wherein the TMU is further configured to:
train a first machine learning model based on (i) a first set of prior two-dimensional image data and (ii) a first set of corresponding trailer statuses, and a second machine learning model based on (i) a second set of prior two-dimensional image data and (ii) a second set of corresponding trailer statuses;
apply the first machine learning model to the two-dimensional image data to generate a preliminary trailer status, wherein the preliminary trailer status indicates one of (i) a closed trailer door, (ii) an other trailer status, or (iii) an unknown trailer status; and
in response to generating a preliminary trailer status indicating the other trailer status, apply the second machine learning model to the two-dimensional image data to generate the second trailer status.
10. The system of claim 9 , wherein the TMU is further configured to:
in response to generating the preliminary trailer status indicating the other trailer status, crop the two-dimensional image data; and
apply the second machine learning model to the cropped two-dimensional image data to generate the second trailer status.
11. The system of claim 9 , wherein the first set of corresponding trailer statuses includes indications that a respective image in the first set of prior two-dimensional image data represents at least one of (i) the closed trailer door, (ii) the other trailer status, or (iii) the unknown trailer status;
wherein the second set of prior two-dimensional image data is a set of prior cropped two-dimensional data; and
wherein the second set of corresponding trailer statuses includes indications that a respective image in the set of prior cropped two-dimensional image data represents at least one of (i) an ajar trailer door, (ii) an open trailer door, or (iii) a parking lot.