Authors:
Gary Harper, Abby Beizer, Jordan Case, Ramon Reveron
Abstract:
This disclosure proposes a method of determining if a shopping cart has been abandoned using an automated reverse-forward analysis of instore video data.
Background:
Customers often abandon shopping carts within stores, with and without merchandise in them. This leads to potential for spoilage, out of place merchandise, clutter, and general untidiness, in addition to other problems. Current methods of detection rely on waiting long periods of time, guessing, and other inefficient means of determination.
Description:
The area to be monitored should have full video coverage of open areas where carts and customers may visit. Coverage may consist of multiple video feeds if a single camera cannot cover the entire area.
During real-time, each frame of video will be analyzed to detect carts. Whenever a cart without an owner (cart without an owner is defined as a cart without a person in a position that indicates they are pushing it as determined using AI/Machine Learning analysis) is detected, it will be marked as potentially abandoned and enter threshold Level I. Carts will remain in threshold Level I until either the cart regains an owner or a configurable time is exceeded. A Level I cart that exceeds the ownerless time threshold will transition to Level II. Once a cart enters Level II, the automated abandonment detection algorithm will begin analyzing the saved video data starting at the time the cart was first identified as a Level I cart. Video can be analyzed in reverse time sequence by the AI/Machine learning algorithm to find the last time the owner was with the cart. The algorithm will then begin tracking the owner of the cart in forward time sequence. This process will continue until either the owner returns to the cart, where it will no longer be considered a potential abandoned cart or until the owner (1) leaves the store, (2) gets a different cart, or (3) some time elapses that indicates the owner no longer is interested in using the cart. At this time, the cart will transition to Level III and raise an alert.
Once a cart has reached Level III, a store associate can recover the cart or mark it as an exception. An example of an exception may be a cart that is temporarily used for display/storage of store merchandise.
Usages:
1) Minimize the time abandoned food that can be spoiled is not refrigerated.
2) Minimize time merchandise is not stocked in the correct location.
3) Minimize clutter within aisles.
Enabling Technology:
It is required that the store have complete video coverage and if multiple cameras are used, a system of overlapping video is employed so that the same cart that appears on multiple videos feeds is not mistaken to be multiple carts. A mapping of the store must be utilized so that as a cart or person moves from the area of one camera to another, it is apparent where the person is coming from or going to so that tracking within the store can be accomplished in a seamless and unambiguous manner.
TGCS Reference 2498
