System and method for stock inventory management
A method of inventory management involves steps to obtain a real-time image of a scene using a sensor, filter the real-time image to delineate portions of the scene, predict objects in the filtered real-time image, identify portions of the objects, and classify the objects using a trained model and the identified portions. The method transmits the real-time image to a remote site configured to generate updates to the trained model. The trained model is sent via over-the-air updates to the trained model. A method of image detection and training involves steps to receive image information of a scene, filter the image information to specify delineated portions of the scene, label portions of the image information, train a convolutional neural network to identify features of the labeled portions, extract the features of the labeled portions, and train a model based on the extracted features to be used in connection with real-time object detection.
1. A method of inventory management, the method comprising:
responsive to detecting a transport vehicle and determining at least one object of one or more objects inserted into the transport vehicle, obtaining, by a central processing unit (CPU), a real-time image of a scene including the one or more objects corresponding to aircraft stock using a sensor;
filtering, by the CPU, the real-time image to delineate portions of the scene;
predicting, by the CPU, both a location and an initial type of at least one object of the one or more objects corresponding to the aircraft stock in the filtered real-time image by analyzing the filtered real-time image;
identifying, by the CPU, at least one portion of the at least one object of the initial type predicted; and
classifying, by the CPU, a final type of the at least one object using a trained model based at least in part on the location, the initial type predicted, and the at least one identified portion.
2. The method of claim 1 , further comprising:
transmitting, by the CPU, the real-time image to a remote site configured to generate updates to the trained model.
3. The method of claim 1 , further comprising:
receiving, by the CPU, over-the-air updates to the trained model from a remote site.
4. The method of claim 1 , wherein the at least one object is predicted by using a mask to detect features of the at least one object.
5. The method of claim 1 , wherein the at least one portion of the at least one object is a pixel.
6. The method of claim 1 , wherein the at least one object is classified as the transport vehicle.
7. The method of claim 1 , wherein the at least one object is classified as an item inserted at a location.
8. The method of claim 1 , wherein the at least one object is classified as an item removed from a location.
9. The method of claim 1 , wherein the at least one object is classified as an aircraft catering item.
10. The method of claim 1 , wherein at least one object is classified using the trained model to manage inventory more efficiently and reduce waste and cost through optimized catering loading plans.
11. A stock management system comprising:
a sensor configured to receive a real-time image of a scene;
a wireless communication device configured to transmit and receive wireless data;
a vision recognition system comprising a central processing unit (CPU), the CPU configured to utilize the sensor to, responsive to detecting a transport vehicle and determining at least one object of one or more objects inserted into the transport vehicle, obtain a real-time image of the scene including the one or more objects corresponding to aircraft stock, filter the real-time image to delineate portions of the scene, predict both a location and an initial type of at least one object of the one or more objects corresponding to the aircraft stock in the filtered real-time image by analyzing the filtered real-time image, identify at least one portion of the at least one object of the initial type predicted, and classify a final type of the at least one object using a trained model based at least in part on the location, the initial type predicted, and the at least one identified portion; and
a remote machine learning system configured to generate updates to the trained model using the wireless communication device.
12. The stock management system of claim 11 , wherein the remote machine learning system is configured to receive image information of a scene, filter the image information to specify delineated portions of the scene, label at least one portion of the image information, train a convolutional neural network to identify features of the at least one portion of the image information, extract the features of the at least one portion of the image information, and train a model based on the extracted features to be used in connection with real-time object detection.
13. The stock management system of claim 11 , further comprising:
a security system module configured to provide security for at least one of the wireless data, the real-time image, the at least one object, the at least one portion of the at least one object, and the trained model.
14. The stock management system of claim 13 , wherein the security system module is a hardware security module.
15. The stock management system of claim 13 , wherein the security system module is a software security module.