Training of vision detection systems using rfid tags
AssigneeImpinj, Inc.
InventorChristopher J. DIORIO
Life of this patent
Estimating the term…
Abstract
A recycling-material sorting system includes an image capture system, an RFID reader system, and a vision detection system. A controller receives item identifier information from an RFID IC associated with a recyclable item through the RFID reader system. If the controller is able to derive recycling information for the recyclable item, it is used to sort the recyclable item. The item identifier information is further used to train a vision detection system regardless of whether the recycling information can be derived for the recyclable item or not.
Claims10
A recycling-material sorting system comprising:
- an image capture system configured to capture images of potentially recyclable items;
- a radio frequency identification (RFID) reader system configured to communicate with RFID integrated circuits (ICs) associated with the items;
- a vision detection system configured to identify item classes from the images of the items based on at least one supervised learning algorithm;
- and
- a controller configured to:
- receive, from the RFID reader system, a first item identifier for a first potentially recyclable item, wherein:
- the RFID reader system retrieves the first item identifier from an RFID IC associated with the first item, and
- the first item identifier indicates at least a first item class for the first item;
- receive, from the image capture system, a first image of the first item captured by the image capture system;
- provide training set data including the first item class and the first image to the vision detection system for training the supervised learning algorithm;
- receive, from the image capture system, a second image of a second potentially recyclable item;
- provide the second image to the vision detection system, wherein the vision detection system identifies the first item class from the second image and provides the identified first item class to the controller;
- and
- use at least the identified first item class to determine recycling information for the second item.
A method for a recycling-material sorting system, the method comprising:
- capturing images of potentially recyclable items through an image capture system;
- communicating with radio frequency identification (RFID) integrated circuits (ICs) associated with the potentially recyclable items through an RFID reader system;
- identifying item classes from the images of the potentially recyclable items based on at least one supervised learning algorithm through a vision detection system;
- and
- through a controller:
- receiving, from the RFID reader system, a first item identifier for a first potentially recyclable item, wherein:
- the RFID reader system retrieves the first item identifier from an RFID IC associated with the first item, and
- the first item identifier indicates at least a first item class for the first item;
- receiving, from the image capture system, a first image of the first item captured by the image capture system;
- providing training set data including the first item class and the first image to the vision detection system for training the supervised learning algorithm;
- receiving, from the image capture system, a second image of a second potentially recyclable item;
- providing the second image to the vision detection system, wherein the vision detection system identifies the first item class from the second image and provides the identified first item class to the controller;
- and
- using at least the identified first item class to determine recycling information for the second item.
Dependencies are inferred from the claim text.