Drone based automated yard check
Methods and systems for providing mechanisms for automated inventory control are provided. In embodiments, an operational workflow for providing automated inventory control includes automated data collection and automated image data analysis. The automated data collection includes capturing image data for a storage facility by a capturing device (e.g., an unmanned aerial system (UAS)). The automated image data analysis includes functionality to detect objects (e.g., containers, trailers, empty slots, and/or other objects) appearing in the image data (e.g., using a machine learning (ML) model), to identify the objects in the image data (e.g., using the ML model), to inspect the objects (e.g., including determining a condition and/or location of the objects) based on the collected image data and metadata associated with the image capturing device and correlated to the image data (e.g., using an advance mathematical rule-based analysis), and/or to generate results that may be used by an inventory management system.
1. An automated inventory control system, comprising:
a processor; and
a memory operably coupled to the processor and storing processor-readable code that, when executed by the processor, is configured to:
detect at least one object in at least one image of a plurality of input images received from an image capturing device;
inspect each object detected in each image of the set of images; and
determine a heading of an observation of an object in an image by:
determining a horizontal pixel location of a center of an object detected in the image;
determining a relative angle from the image capturing device to the object;
determining an object heading based on a heading of the image capturing device and the relative angle;
generating, for each object detection in each image of the set of images, a line that is based on the observation heading for each object detection and a location of the image capturing device at the time the image was captured;
determining a location of the particular unique object by calculating a mean of all line intersections in a region of intersections for all observations of the object, and
discarding line pair combinations that have an intersection angle less than a predetermined threshold.
2. The system of claim 1 , wherein the processor-readable code is further configured to correlate metadata associated with at least one frame to the at least one image of the plurality of input images.
3. The system of claim 1 , wherein the processor-readable code is further configured to determine, based at least on the at least one image, an identification of the at least one object detected in the at least one image.
4. The system of claim 1 , wherein inspecting an object includes determining a condition and location of the object.
5. The system of claim 4 , wherein determining a condition of each object detected in each image of the set of images includes applying a trained machine learning (ML) model to each image to detect a condition of each object detected in each image.
6. The system of claim 5 , wherein the trained ML model is configured to analyze an image and determine any damage to the at least one detected object, whether the at least one detected object is open or closed, whether the at least one detected object is positioned correctly, or whether the at least one detected object is an empty slot.
7. The system of claim 1 , wherein the relative angle is determined based on the following equation:
β
=
P
o
W
×
θ
F
O
V
,
where θ FOV represents a field of view of the image capturing device as an angle and W represent the width of the image in which the object is detected in pixels.
8. The system of claim 7 , wherein the object heading is determined based on the following equation:
θ Heading Object =θ Heading Camera −β.
9. A method of determining a heading of an observation of an object in an image including:
determining, via a location detector, a horizontal pixel location of a center of an object detected in an image received from an image capturing device;
determining a relative angle from the image capturing device to the object;
determining an object heading based on a heading of the image capturing device and the relative angle;
generating, for each object detection in each image of the set of images, a line that is based on the observation heading for each object detection and a location of the image capturing device at the time the image was captured; and
determining a location of the particular unique object by calculating a mean of all line intersections in a region of intersections for all observations of the object, and
discarding line pair combinations that have an intersection angle less than a predetermined threshold.
10. The method of claim 9 , wherein the image processor correlates metadata associated with at least one frame to the image.
11. The method of claim 9 , wherein the image processor determines, based at least on the at least one image, an identification of the at least one object detected in the at least one image.
12. The method of claim 9 , wherein inspecting an object includes determining a condition and location of the object.
13. The method of claim 12 , wherein determining a condition of each object detected in each image of the set of images includes applying a trained machine learning (ML) model to each image to detect a condition of each object detected in each image.
14. The method of claim 13 , wherein the trained ML model is configured to analyze an image and determine any damage to the at least one detected object, whether the at least one detected object is open or closed, whether the at least one detected object is positioned correctly, or whether the at least one detected object is an empty slot.
15. The method of claim 9 , wherein the relative angle is determined based on the following equation:
β
=
P
o
W
×
θ
F
O
V
,
where θ FOV represents a field of view of the image capturing device as an angle and W represent the width of the image in which the object is detected in pixels.
16. The method of claim 15 , wherein the object heading is determined based on the following equation:
θ Heading Object =θ Heading Camera −β.