Granted Patent B2

US 12,361,374 · App. 17/989,577 · Granted Jul 15, 2025

Inventors: Job Varughese Philip (Mumbai, IN); Rohit Shekhar Pingulkar (Mumbai, IN); Gourav Raju Bhure (Nagpur, IN); Rajesh Jagnarayan Roy (Mumbai, IN)

Assignee: ASSERT SECURE TECH PVT. LIMITED

G06Q10/087

Abstract

Claims (37)

1. A system for tracking an inventory inside a warehouse with put-away accuracy using machine learning models, wherein the system comprises,

an unmanned aerial vehicle (UAV) that is configured to capture a plurality of media contents of a pre-defined space within the warehouse when in operation, using an image capturing device, wherein the UAV is pre-programmed to (i) identify a plurality of rack bays, (ii) define a moving path within the warehouse to reach a location of the plurality of rack bays, and (iii) move from one rack bay to another automatically based on a layout of the warehouse and coordinates of the plurality of rack bays; and

an inventory tracking unit that is communicatively connected with the UAV and the inventory tracking unit comprises,

a memory that stores a database and a set of instructions, and

a processor that executes the set of instructions from the memory and is configured to:

receive, by a first machine learning model, the plurality of media contents associated with the pre-defined space in real-time from the UAV to determine a plurality of inventory items, and the plurality of rack bays associated with the pre-defined space, wherein the pre-defined space comprises the plurality of rack bays and each rack bay comprises a unique rack bay identifier, wherein each rack bay is configured to stock the plurality of inventory items with a pallet identifier that is matched with the unique rack bay identifier of the corresponding rack bay;

train the first machine learning model by correlating historical media contents with historical inventory items, and historical rack bays, wherein the historical media contents are captured in at least one of different lighting conditions comprising at least one of low light conditions, or bright light conditions, or Infra-red (IR) mode;

generate, using the first machine learning model, empty space data by determining an empty space on the plurality of rack bays between the plurality of inventory items when the plurality of inventory items is not detected;

train a second machine learning model by correlating the historical inventory items, and the historical rack bays that are associated with at least one identifier with historical optical characters of the at least one identifier in the historical media contents;

identify, using the second machine learning model, at least one of (i) the unique rack bay identifier associated with each rack bay, and (ii) the pallet identifier associated with the plurality of inventory items by recognizing one or more optical characters from the plurality of inventory items, and the plurality of rack bays that are detected by the first machine learning model;

validate whether the pallet identifier associated with the plurality of inventory items that is identified is matched with the unique rack bay identifier of the corresponding rack bay; and

send an empty space alert along with the empty space data, if the empty space on the plurality of rack bays is determined, thus enabling a user to track the inventory inside the warehouse with the put-away accuracy, wherein the put-away accuracy is a percentage of a total number of the plurality of inventory items put away correctly and a total number of items put away.

2. The system of claim 1 , wherein the processor is configured to generate (i) mismatched data, when the pallet identifier is not matched with the corresponding unique rack bay identifier, and (ii) matched data, when the pallet identifier is matched with the corresponding unique rack bay identifier.

3. The system of claim 1 , wherein the processor is configured to (i) count the total number of the plurality of inventory items in each rack bay to generate data on the total number of items put away, and (ii) record the inventory to generate recorded inventory details by reading an inventory identifier on the plurality of inventory items using the second machine learning model.

4. The system of claim 2 , wherein the processor is configured to send a mismatch alert along with the mismatched data, if the pallet identifier associated with the plurality of inventory items is not matched with the unique rack identifier of the corresponding rack bay to a user device associated with the user.

5. The system of claim 2 , wherein the processor is further configured to communicate the empty space data, the mismatched data, the matched data, the total number of the plurality of inventory items and the recorded inventory details to at least one of (i) a cloud server or (ii) the user device through a network.

6. A method of tracking an inventory inside a warehouse with put-away accuracy using machine learning models comprising,

capturing, using an image capturing device from an unmanned aerial vehicle (UAV), a plurality of media contents of a pre-defined space within the warehouse, wherein the UAV is pre-programmed to (i) identify a plurality of rack bays, (ii) define a moving path within the warehouse to reach a location of the plurality of rack bays, and (iii) move from one rack bay to another automatically based on a layout of the warehouse and coordinates of the plurality of rack bays;

receiving, by a processor of an inventory tracking unit, the plurality of media contents associated with the pre-defined space in real-time from the UAV to determine a plurality of inventory items, and the plurality of rack bays associated with the pre-defined space, wherein the pre-defined space comprises the plurality of rack bays and each rack bay comprises a unique rack bay identifier, wherein each rack bay is configured to stock the plurality of inventory items with a pallet identifier that is matched with the unique rack bay identifier of the corresponding rack bay;

training a first machine learning model, by correlating historical media contents with historical inventory items, and historical rack bays, wherein the historical media contents are captured in at least one of different lighting conditions comprising at least one of low light conditions, or bright light conditions, or Infra-red (IR) mode;

generating, using the first machine learning model, empty space data by determining an empty space on the plurality of rack bays between the plurality of inventory items when the plurality of inventory items is not detected;

training a second machine learning model by correlating the historical inventory items, and the historical rack bays that are associated with at least one identifier with historical optical characters of the at least one identifier in the historical media contents;

identifying, using the second machine learning model, at least one of (i) the unique rack bay identifier associated with each rack bay, and (ii) the pallet identifier associated with the plurality of inventory items by recognizing one or more optical characters from the plurality of inventory items, and the plurality of rack bays that are detected by the first machine learning model;

validating, by the processor, whether the pallet identifier associated with the plurality of inventory items that is identified is matched with the unique rack bay identifier of the corresponding rack bay; and

sending, by the processor, an empty space alert along with the empty space data, if the empty space on the plurality of rack bays is determined, thus enabling a user to track the inventory inside the warehouse with the put-away accuracy, wherein the put-away accuracy is a percentage of a total number of inventory items put away correctly and a total number of items put away.

7. The method of claim 6 , wherein the method comprising (i) counting the total number of the plurality of inventory items in each rack bay to generate data on the total number of items, and (ii) recording the inventory to generate recorded inventory details by the processor when reading an inventory identifier on the plurality of inventory items using the second machine learning model.

8. The method of claim 6 , wherein the method comprising communicating, by the processor, (i) the empty space data, (ii) mismatched data that is generated, if the pallet identifier associated with the plurality of inventory items is not matched, (iii) matched data that is generated, if the pallet identifier associated with the plurality of inventory items is matched, (iv) the data on the total number of the plurality of inventory items and (v) the recorded inventory details to at least one of (i) a cloud server or (ii) a user device through a network.

9. A non-transitory computer-readable storage medium storing a sequence of instructions, which when executed by a processor, causes performing a method of tracking an inventory inside a warehouse with put-away accuracy using machine learning models comprising, capturing, using an image capturing device from an unmanned aerial vehicle (UAV), a plurality of media contents of a pre-defined space within the warehouse, wherein the UAV is pre-programmed to (i) identify a plurality of rack bays, (ii) define a moving path within the warehouse to reach a location of the plurality of rack bays, and (iii) move from one rack bay to another automatically based on a layout of the warehouse and coordinates of the plurality of rack bays;

receiving, by a processor of an inventory tracking unit, the plurality of media contents associated with the pre-defined space in real-time from the UAV to determine a plurality of inventory items, and the plurality of rack bays associated with the pre-defined space, wherein the pre-defined space comprises the plurality of rack bays and each rack bay comprises a unique rack bay identifier, wherein each rack bay is configured to stock the plurality of inventory items with a pallet identifier that is matched with the unique rack bay identifier of the corresponding rack bay;

training a first machine learning model, by correlating historical media contents with historical inventory items, and historical rack bays, wherein the historical media contents are captured in at least one of different lighting conditions comprising at least one of low light conditions, or bright light conditions, or Infra-red (IR) mode;

generating, using the first machine learning model, empty space data by determining an empty space on the plurality of rack bays between the plurality of inventory items when the plurality of inventory items is not detected;

training a second machine learning model by correlating the historical inventory items, and the historical rack bays that are associated with at least one identifier with historical optical characters of the at least one identifier in the historical media contents;

identifying, using the second machine learning model, at least one of (i) the unique rack bay identifier associated with each rack bay, and (ii) the pallet identifier associated with the plurality of inventory items by recognizing one or more optical characters from the plurality of inventory items, and the plurality of rack bays that are detected by the first machine learning model;

validating, by the processor, whether the pallet identifier associated with the plurality of inventory items that is identified is matched with the unique rack bay identifier of the corresponding rack bay; and

sending, by the processor, an empty space alert along with the empty space data, if the empty space on the plurality of rack bays is determined, thus enabling a user to track the inventory inside the warehouse with the put-away accuracy, wherein the put-away accuracy is a percentage of a total number of inventory items put away correctly and a total number of items put away.

10. The non-transitory computer-readable storage medium storing a sequence of instructions of claim 9 , wherein the method comprising (i) counting the total number of the plurality of inventory items in each rack bay to generate data on the total number of items, and (ii) recording the inventory to generate recorded inventory details by the processor when reading an inventory identifier on the plurality of inventory items using the second machine learning model.

11. The non-transitory computer-readable storage medium storing a sequence of instructions of claim 9 , wherein the method comprising communicating, by the processor, (i) the empty space data, (ii) mismatched data that is generated, if the pallet identifier associated with the plurality of inventory items is not matched, (iii) matched data that is generated, if the pallet identifier associated with the plurality of inventory items is matched, (iv) the data on the total number of the plurality of inventory items and (v) the recorded inventory details to at least one of (i) a cloud server or (ii) a user device through a network.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (19)

Cited By (1)

US 12,703,614